Ten cities including Beijing, Shijiazhuang, and Jinan in northern China, Lanzhou in northwestern China, Zhengzhou and Wuhan in central China, Shanghai in eastern China, Chengdu and Guiyang in southwestern China, and Guangzhou in southern China were selected for measuring TGM concentrations and isotopic compositions (Fig. S1 in the Supplement). These cities are located in different geographical regions of China, which were potentially characterized by specific source emission patterns, climate, and atmospheric chemistry. The designated investigations in these cities may therefore provide comprehensive information on the variations of TGM concentrations and isotopic compositions in the megacities of China and help to explore the major factors influencing the atmospheric TGM in Chinese cities. Site locations, information about the 10 cities and sampling periods are given in Table S1 in the Supplement. Briefly, these cities have populations of 3.75 to 21.54 million in urban areas. Fractions of the urban and construction land area out of the total land area of a 1∘ × 1∘ grid surrounding the sampling sites ranged from 4.9 % to 41.1 % (mean = 22.5 %), whereas the remaining land surfaces are mainly croplands, barren lands, open grassland, open shrublands and open forests (range from 35.6 % to 82.3 % with a mean of 57.5 %) (Fig. S2 and Table S1 in the Supplement). In each of the 10 cities, one sampling site was selected for measuring TGM concentrations and isotopic compositions. The sampling sites are generally located in heavily commercial and residential areas in all the cities and with no major industrial Hg emission sources within 2 km of the sampling sites. All the measurements were conducted on building roofs at elevations of >10 m. In order to investigate the shift of GEM isotopic compositions in urban areas relative to that in remote areas due to local urbanized emissions, the same type of measurements were also conducted at the rural Waliguan Baseline Observatory in northwestern China (Mt. Waliguan), which belongs to the World Meteorological Organization's (WMO) Global Atmospheric Watch (GAW) network (Fig. S1 in the Supplement).

In this study, chlorine-impregnated activated carbon (CLC; 0.5 g) traps were used to collect atmospheric TGM samples (Fu et al., 2014). A schematic diagram of the sampling system is shown in Fig. S3 in the Supplement. Briefly, particles in ambient air were first removed using a Teflon filter (47 mm diameter; 0.2 µm pore size) at the inlet of the sampling system, and then ambient TGM were trapped onto the CLC traps at a flow rate of ∼2.5 L min-1 using a Teflon-coated mini-diaphragm vacuum pump (N89 KTDC, KNF Inc.). The sampling flow rate was adjusted using a needle valve installed at the outlet of the vacuum pump. The inlet of the sampling system was about 1.5 m above the surface. Daily (24 h) continuous sampling of TGM at each urban site lasted for approximately 1 week in the winter and summer of 2018, respectively (daily samples; Table S1 in the Supplement). The wintertime samplings were conducted from 5 to 15 January 2018 simultaneously in Beijing, Shijiazhuang, Jinan, Lanzhou and Zhengzhou, and from 18 to 27 January 2018 simultaneously in Shanghai, Chengdu, Wuhan, Guiyang and Guangzhou. The summertime samplings were conducted from 29 June to 7 July 2018 simultaneously in Shijiazhuang, Jinan, Zhengzhou, Guiyang and Guangzhou, and from 27 July to 10 August 2018 simultaneously in Beijing, Lanzhou, Shanghai, Wuhan and Chengdu. TGM samples were also continuously collected from 19 November 2014 to 19 February 2015 at Mt. Waliguan with a sampling duration of 10 d. After field sampling, CLC traps were sealed carefully and kept in a sealed polypropylene crisper before sample processing for Hg isotope analysis.

GOM concentrations are generally elevated in Chinese urban areas due to local primary anthropogenic emissions (Fu et al., 2015). Previous studies showed that GOM measured using a Tekran 2537/1130/1135 system on average accounted for 0.37 % to 0.50 % of TGM in Guiyang, Beijing and Shanghai, China, while the daily GOM fractions in TGM ranged from 0.04 % to 1.58 % in Guiyang (Table S2 in the Supplement) (Fu et al., 2011; Duan et al., 2017; Zhang et al., 2019). It is likely that the Tekran system could underestimate GOM concentrations by approximately 3-fold with respect to that measured by other recently developed methods (e.g., cation exchange membranes (CEMs) or nylon membranes) (Huang et al., 2013; Gustin et al., 2015, 2019). To date, GOM has not been measured by CEM or nylon membranes in Chinese urban areas. If adjusting GOM concentrations by a factor of 3, the abovementioned mean GOM fractions would be increased to 1.1 %–1.5 % in Guiyang, Beijing and Shanghai, and these values are similar to those observed in Reno, Nevada, USA, based on the CEM method (mean GOM fraction of 2.7 %) (Gustin et al., 2019). Mean δ202Hg and Δ199Hg of TGM in urban areas of this study ranged from -0.96 ‰ to -0.24 ‰ and -0.12 ‰ to -0.01 ‰, respectively. Assuming that the isotope composition of GOM resemble those of primary anthropogenic emissions (e.g., δ202Hg = -0.77 ‰, Δ199Hg = -0.06 ‰) (R. Y. Sun et al., 2016), a maximum GOM fraction of TGM (5 %) would lead to negligible shifts in TGM δ202Hg (-0.03 ‰ to 0.01 ‰) and Δ199Hg (-0.003 ‰ to 0.003 ‰). Therefore, TGM isotopic composition measured in this study would not be biased significantly by GOM compounds. The use of the terms “TGM” and “GEM” interchangeably in this and previous studies would not significantly confound the intercomparison of isotopic composition.

Before the analysis of Hg concentration and isotopic composition, TGM collected on CLC traps were preconcentrated into 5 mL of 2HNO3 / 1HCl mixed acid solution (40 %) following previous studies (Biswas et al., 2008; Sun et al., 2013; Fu et al., 2014). Trapping solution Hg concentrations were measured by a Tekran 2500 Hg analyzer following US EPA Method 1631 (USEPA, 2002). TGM concentrations of the samples were calculated using Eq. (1): 1TGM=C×VsolutionVgas, where TGM is the atmospheric TGM concentration in ng m-3, C is the Hg concentration in the trap solution in ng mL-1, Vsolution is the volume of the trap solution in mL and Vgas is the cumulative sampling air volume in m3. Full procedural blanks of field sampling and preconcentration were measured at each sampling site and in each season by combustion of sealed field CLC traps (containing 0.5 g CLC) prepared before field sampling. The mean Hg concentration in these sealed field blanks was 0.20 ± 0.09 ng (1σ, n=27; Table S3 in the Supplement), which was negligible (<5 %) compared to the Hg in trapping solutions of samples. Breakthrough tests showed that 96.7 % to 99.6 % (mean = 98.9 ± 0.9 %, 1σ, n=10) of TGM in ambient air could be collected by the CLC traps in our experiment setting (Table S3 in the Supplement). Recoveries of the preconcentration were tested by combustion of lichen CRM (BCR 482), which showed a mean value of 92.5 ± 3.9 % (1σ, n=6; Table S3 in the Supplement). Standard additions of Hg0 vapor (5 to 25 ng; produced by SnCl2 reduction of diluted NIST 3133 solutions) to CLC traps at the 2.5 L min-1 sampling flow rate showed a mean recovery of 93.2 ± 11.8 % (1σ, n=11; Table S3 in the Supplement) for the sampling and preconcentration method. These tests indicate that the above method is reliable and efficient for measuring TGM concentrations and isotopic compositions.

Prior to isotope analysis, the concentrations of Hg in trap solution were diluted to 0.5 or 1.0 ng mL-1 using the 2HNO3 / 1HCl mixed acid solution (20 %). Isotope ratios of Hg in diluted trap solutions were measured by cold-vapor multicollector inductively coupled plasma mass spectrometry (CV-MC-ICPMS) using a Nu Plasma (Nu Instruments) and a Neptune (Thermo Fisher Scientific) in the Institute of Geochemistry, CAS (Guiyang, China) (Fu et al., 2019). TGM isotopic compositions were calculated following Eqs. (2) and (3) (Blum and Bergquist, 2007): 2δxxxHgTGM‰=xxxHg198HgsamplexxxHg198HgNIST 3133-1×1000, where δxxxHgTGM are the MDF signatures of TGM in per mil (‰), xxx is the mass number of Hg isotopes (199, 200, 201, 202, and 204), (xxxHg / 198Hg)sample is the isotope ratios for TGM samples and (xxxHg / 198Hg)NIST 3133 is the isotope ratios for the bracketing NIST 3133 standard (concentrations matched within 10 % of the sample trapping solution Hg concentrations). 3ΔxxxHgTGM‰=δxxxHgTGM-β×δ202HgTGM, where ΔxxxHgTGM are the MIF signatures of TGM isotopes 199Hg, 200Hg, 201Hg and 204Hg in per mil (‰), and β values are 0.252, 0.5024, 0.752 and 1.493 for isotopes 199Hg, 200Hg, 201Hg and 204Hg, respectively (Blum and Bergquist, 2007).

Isotopic compositions of NIST 3177 Hg standard (n=30), lichen CRM (BCR 482, n=6) and standard additions of NIST 3133 Hg to CLC traps (n=11) (Table S3 in the Supplement) were analyzed periodically during TGM isotope analysis, and the results were consistent with previously reported values or the original values of the NIST 3133 Hg standard (Table S3 in the Supplement) (Enrico et al., 2016; G. Sun et al., 2016; Blum and Johnson, 2017). In the present study, we report the analytical uncertainties (2σ) of TGM isotopic compositions as the 2σ values of the sample replicates when they are higher than the 2σ values of standard addition of NIST 3133 Hg to CLC traps. When the 2σ values of the sample replicates were lower than the standard additions of NIST 3133 Hg to CLC traps, 2σ values of the standard additions of NIST 3133 Hg to CLC traps were used to represent the 2σ of TGM isotopic compositions.

Data for concentrations of ozone (O3) and carbon monoxide (CO) during the sampling periods were extracted from national air quality monitoring stations (http://106.37.208.233:20035/, last access: 8 February 2020) located within 1.5 km of the sites, with the exception of the sampling site in Guangzhou (4.1 km). Normalized difference vegetation index (NDVI) around the sampling sites (1∘ × 1∘) was obtained from the NASA Earth Observations (NEO, https://neo.sci.gsfc.nasa.gov/, last access: 30 March 2020).

In order to investigate the effect of soil emissions on the variations in TGM concentrations and isotopic compositions, GEM exchange flux between soil and atmosphere at 1∘ × 1∘ resolution at each sampling site in July and January were extracted from the gridded land surface emission inventory in China simulated for 2013, which has a spatial resolution of ∼36 km and a monthly temporal resolution (Fig. S1 in the Supplement) (Wang et al., 2016). This model established a new scheme for estimating soil–atmosphere GEM flux, which has taken into account the effect of photochemical and nonphotochemical reduction of Hg(II) in soil, diffusion of Hg0 from soil to atmosphere as well as the temperature, moisture, organic matter contents, PH, Hg concentration, bulk density and land cover of soils, etc. (for more detail see Wang et al., 2016). Note that the simulated surface emission inventory does not include GEM emissions from pavement, building surfaces and indoor Hg-containing products. These sources are in close proximity to the sampling sites (Fig. S2 in the Supplement), and their effect is also interpreted in Sect. 3.4.

Daily isotopic compositions of GEM emitted from hillslope barren soil in Guiyang (114.269∘ E, 30.488∘ N) and from agricultural soil in Wuhan (114.269∘ E, 30.488∘ N) from 29 July to 3 August 2019 and from 24 to 27 August 2019, respectively, obtained by measuring the GEM isotopic compositions at the inlet and outlet of a dynamic flux chamber, was followed by a calculation based on the binary mixing model (Eqs. 4 and 5): 4δxxxGEMemission=δxxxGEMoutlet×GEMoutlet-δxxxGEMinlet×GEMinlet÷GEMoutlet-GEMinlet5ΔxxxGEMemission=ΔxxxGEMoutlet×GEMoutlet-ΔxxxGEMinlet×GEMinlet÷GEMoutlet-GEMinlet, where xxx corresponds to the mass number of Hg isotopes (199, 200, 201, 202 (not for the MIF signature) and 204), δxxxGEMoutlet and ΔxxxGEMoutlet are the MDF and MIF values of GEM at the outlet, respectively, δxxxGEMinlet and ΔxxxGEMinlet are the MDF and MIF values of GEM at the inlet, respectively, and GEMoutlet and GEMinlet are the GEM concentrations measured at the outlet and inlet, respectively.

Linear regression analysis was performed with IBM SPSS Statistics using the forced entry method.

Mean TGM concentrations at the urban sites during the study periods ranged from 2.34 to 4.56 ng m-3 (n=10) with a mean (±1σ) of 3.08 ± 0.79 ng m-3 (Fig. 1). These values were 1.5 to 3.0 times higher than the mean background value of 1.51 ng m-3 in 2014 in the Northern Hemisphere obtained from the Global Mercury Observation System (GMOS) (Sprovieri et al., 2016) and 1.2 to 2.4 times higher than the mean value of 1.94 ± 0.64 (±1σ) ng m-3 at urban sites in North America and Europe (Mao et al., 2016). Mean TGM concentrations observed at some urban sites were, however, 44 %–55 % lower than previously reported mean values for earlier years, e.g., 4.19 ng m-3 in Shanghai in 2014, 8.88 ng m-3 in Guiyang in 2010 and 4.60 ng m-3 in Guangzhou in 2011 (Chen et al., 2013; Fu and Feng, 2015; Duan et al., 2017), likely due to a combination of several factors such as decreased anthropogenic emissions (K. Y. Liu et al., 2019), different sampling locations even inside the same city and different sampling times and durations of the year. The declining TGM concentration (by 40 %) in recent years (2014–2016) has indeed been reported in Chongming Island, Shanghai (Tang et al., 2018), which has been mostly attributed to reduced anthropogenic Hg emissions in China. Such emission reductions would impact more on urban than rural areas in atmospheric TGM (K. Y. Liu et al., 2019).

Figure 2 shows the isotopic compositions of daily TGM samples collected at the 10 urban sites and one rural site. Large variations in daily TGM isotopic compositions were observed with values ranging from -1.68 ‰ to 0.63 ‰ for δ202Hg and from -0.23 ‰ to 0.10 ‰ for Δ199Hg (Fig. 2 and Table S4 in the Supplement). Mean TGM δ202Hg values were the lowest in Guiyang (-0.96 ± 0.42 ‰, 1σ), Lanzhou (-0.70 ± 0.35 ‰, 1σ) and Chengdu (-0.68 ± 0.44 ‰, 1σ) in southwestern and northwestern China, followed by Wuhan (mean = -0.68 ± 0.23 ‰, 1σ) and Zhengzhou (mean = -0.55 ± 0.24 ‰, 1σ) in central China, Shijiazhuang (-0.54 ± 0.44 ‰, 1σ) and Jinan (-0.50 ± 0.42 ‰, 1σ) in northern China, coastal Guangzhou (-0.44 ± 0.17 ‰, 1σ) in southern China, coastal Shanghai (-0.32 ± 0.35 ‰, 1σ) in eastern China, and was the highest in Beijing (-0.24 ± 0.24 ‰, 1σ) in northern China (Fig. 1 and Table S5 in the Supplement). Much smaller spatial variations were seen in mean TGM Δ199Hg than TGM δ202Hg. The highest mean TGM Δ199Hg were observed in Guiyang (-0.01 ± 0.06 ‰, 1σ) and Chengdu (-0.01 ± 0.03 ‰, 1σ), whereas values at the other urban sites ranged from -0.12 ‰ to -0.04 ‰ (n=8). Mean TGM δ202Hg (-0.16 ± 0.16 ‰, 1σ) (or Δ199Hg = -0.10 ± 0.04 ‰, 1σ) values measured at rural Mt. Waliguan in winter were higher (or lower) than that in most cities in winter, with the exception of δ202Hg in Beijing and Shanghai (means = -0.09 ‰ to -0.07 ‰, n=2) and Δ199Hg in Shijiazhuang, Jinan, and Shanghai (means = -0.17 ‰ to -0.11 ‰, n=3), where TGM concentrations were low (means = 1.88 to 2.12 ng m-3) and comparable to that at rural Mt. Waliguan (Table S5 in the Supplement). Mean TGM Δ200Hg values at the urban sites were all indistinguishable from zero (-0.03 ‰ to 0.02 ‰, n=10; Table S5 in the Supplement), a phenomenon that is similar to previous observations in urban areas in China and the USA (means = -0.01 ‰ to 0.01 ‰ , n=4) (Gratz et al., 2010; Yu et al., 2016; Xu et al., 2017). Therefore, we do not further interpret the MIF of even-mass Hg isotopes in this study.

Mean values of TGM δ202Hg and Δ199Hg in this study were similar to those reported at urban sites in China in previous studies, e.g., negative δ202Hg (means = -0.73 ‰ to -0.08 ‰, n=3) and close to zero Δ199Hg (means = -0.03 ‰ to 0.04 ‰, n=3) in Beijing, Xi'an and Guiyang (Yu et al., 2016; Xu et al., 2017). On the other hand, mean TGM δ202Hg values in this study were 0.44 ‰ to 1.60 ‰ lower than the values reported for rural areas of China (mean = 0.20 ± 0.40 ‰, 1σ, n=3) and North America and Europe (mean = 0.71 ± 0.39 ‰, 1σ, n=7), whereas mean TGM Δ199Hg values were 0.02 ‰ to 0.18 ‰ higher than the means in rural areas of China (mean = -0.14 ± 0.05 ‰, 1σ) and North America and Europe (mean = -0.20 ± 0.05 ‰, 1σ) (Fig. 2) (Gratz et al., 2010; Demers et al., 2013, 2015; Enrico et al., 2016; Fu et al., 2016; Obrist et al., 2017; Fu et al., 2018, 2019; Jiskra et al., 2019b). Apparently, atmospheric TGM is isotopically distinguishable between urban and rural sites and between different regions of the world, providing a potentially valuable tracer for understanding the sources and transformations of atmospheric Hg at local, regional and global scales. As shown in Fig. 2, some of the daily TGM isotopic compositions (i.e., δ202Hg and Δ199Hg signatures) fell between the end-member TGM isotopic compositions estimated for anthropogenic TGM emissions and observed from background areas, suggesting mixed influences on TGM isotopic compositions between anthropogenic emissions and background atmospheric pool (Demers et al., 2015; Fu et al., 2016; Xu et al., 2017; Fu et al., 2018). There were, however, many exceptions with daily TGM isotopic compositions outside the abovementioned range, e.g., with δ202Hg lower than -0.75 ‰ or Δ199Hg higher than -0.01 ‰ (Fig. 2). Thus, additional sources and environmental processes should have also contributed to the variations in TGM isotopic compositions in urban environments of China.

Mean TGM concentrations at the urban sites (2.37 to 4.56 ng m-3, n=10) were highly elevated compared to the background value (∼1.5 ng m-3) in the Northern Hemisphere (Sprovieri et al., 2016). This could be attributed to local and regional Hg sources including Hg emissions from primary anthropogenic sources, land surfaces (e.g., soil, building, and pavement) and indoor Hg-containing products. As shown in Fig. 2, TGM isotopic compositions in the cities were probably controlled by a binary physical mixing between the regional-scale background and the key end-member sources in the cities, which could be associated with the local and regional emission sources. Here we use a linearized binary physical mixing diagram to estimate the mean isotopic signature of the urbanized source end-members by extrapolating the 1 / TGMmean to zero (where TGM is mostly derived from urbanized sources) (Fig. 3), which showed δ202Hg and Δ199Hg values of approximately -1.16 ± 0.15 ‰ and 0.05 ± 0.02 ‰ (1σ), respectively.

The estimated δ202Hg (or Δ199Hg) for urbanized emissions was much lower (or much higher) than the δ202Hg of -0.26 ‰ (or Δ199Hg of -0.26 ‰) for GEM emitted from CFPPs in China (Tang et al., 2017; H. W. Liu et al., 2019). The isotopic signatures of other anthropogenic emission sectors in China have not been appropriately constrained. R. Y. Sun et al. (2016) estimated a mean δ202Hg of -0.59 ‰ and a mean Δ199Hg of -0.02 ‰ for the global anthropogenic GEM emissions in 2010. Our estimate of δ202Hg and Δ199Hg for the urbanized sources were, however, 0.57 ‰ lower and 0.07 ‰ higher than their predicted value for anthropogenic emissions, respectively. A recent study by Jiskra et al. (2019a) showed highly negative δ202Hg (means = -1.54 ‰ to -1.56 ‰, n=2) and high Δ199Hg values (means = 0.00 ‰ to 0.17 ‰, n=2) for GEM in air impacted by Hg emissions from building surfaces and indoor sources, and these values seemed to support, to some extent, the estimated negative δ202Hg and close to zero Δ199Hg signatures of urbanized sources in the present study.

Soil emissions are a potentially important source of atmospheric TGM in urban areas (Feng et al., 2005; Agnan et al., 2016), and GEM emission fluxes from urban soils were reported to be approximately one order of magnitude higher than that from pavement and building surfaces (Gabriel et al., 2006; Eckley and Branfireun, 2008). The sampling sites in the present study were largely surrounded by cropland and sparsely vegetated soils (Fig. S2 in the Supplement), and it is therefore important to investigate their effects on the variations in TGM concentrations and isotopic compositions. The measured mean GEM emission fluxes from soils in Guiyang and Wuhan in summer were 35.9 ± 32.6 (1σ, n=5) and 9.8 ± 5.3 (1σ, n=3) ng m-2 h-1, respectively. The mean δ202Hg and Δ199Hg values of GEM emitted from soils were -2.16 ± 0.60 ‰ and -0.27 ± 0.15 ‰ (1σ, n=5), respectively, in Guiyang, and were -1.07 ± 0.86 ‰ and -0.01 ± 0.52 ‰ (1σ, n=3), respectively, in Wuhan (Fig. S4 in the Supplement). These values suggest that the isotopic compositions of soil GEM emissions in urban areas of China likely have highly negative δ202Hg values, similar to that of GEM emitted from building surfaces and indoor Hg-containing products (Jiskra et al., 2019a). We thus hypothesize that soil, building surfaces and indoor Hg-containing product emissions contributed to the highly negative TGM δ202Hg values observed in this study. Based on the estimated δ202Hg values of urbanized source end-member (mean = -1.16 ‰), anthropogenic emissions (mean = -0.59 ‰), and GEM emitted from soils, building surfaces and indoor Hg-containing products (mean = -1.57 ‰) in this and previous studies (R. Y. Sun et al., 2016; Jiskra et al., 2019a), we estimate that the contribution of soil, building surfaces and indoor Hg-containing product emissions to the TGM in the 10 cities was approximately equal to that of primary anthropogenic emissions (48 % versus 52 %). We caution that, due to the fact that the isotopic signatures of GEM emitted from many anthropogenic sources and land surfaces in China have not been well constrained, such a preliminary assessment should have large uncertainties. However, our estimate is overall consistent with previous studies on GEM emission fluxes from land surfaces and anthropogenic sources in Chinese urban areas. For example, previous studies on GEM emission fluxes from urban surfaces in China showed a mean value of 83.2 ± 170 ng m-2 h-1 (1σ, n=39) (Fang et al., 2004; Feng et al., 2005; Wang et al., 2006; Fu et al., 2012), which was relatively higher than the mean anthropogenic GEM flux (48.4 ± 48.1 ng m-2 h-1, 1σ, n=10) in the 10 investigated cities (Table S5 in the Supplement) (AMAP/UNEP, 2013). The findings in this and previous studies therefore suggest that soil, building surfaces and indoor Hg-containing product emissions would play an important role in regulating the TGM concentrations and isotopic compositions in urban areas of China, which is further discussed in the following section.

Strong seasonal variations in the mean TGM concentrations and isotopic compositions were observed for most cities (Fig. 4). The mean TGM concentrations and Δ199Hg values were relatively higher in summer than winter in most cities except for the two (Guiyang and Guangzhou) in the low latitudes that showed an opposite trend. On the contrary, the mean TGM δ202Hg showed lower values in summer than winter in all the cities except southernmost Guangzhou, which showed no seasonal difference. The seasonal variations in TGM concentrations and δ202Hg in the present study were consistent with previous findings generated from year-round continuous observations in China, e.g., higher summertime TGM in Beijing and Shanghai (Zhang et al., 2013; Duan et al., 2017), higher wintertime TGM in Guiyang and Guangzhou (Feng et al., 2004; Chen et al., 2013), and lower summer δ202Hg in Guiyang and Xi'an (Yu et al., 2016; Xu et al., 2017).

The summertime higher TGM concentrations observed in most cities in the present study was in contrast to the observations in most rural areas in China as well as in other regions in the Northern Hemisphere, which frequently showed lower TGM or GEM concentrations in summer than in winter (Fu et al., 2015; Mao et al., 2016; Jiskra et al., 2018). Studies on the seasonal variations in TGM or GEM isotopic compositions in rural areas are currently limited. A recent study at rural Mt. Changbai in northeastern China showed higher TGM δ202Hg values in summer than winter (Fu et al., 2019), which is opposite to the seasonal variations in TGM δ202Hg at most urban sites in the present study. Such a summertime lower TGM or GEM level and higher δ202Hg pattern in rural areas should be mainly attributed to increasing atmospheric oxidation and vegetation uptake of GEM as well as decreasing residential coal combustion (Sprovieri et al., 2016; Horowitz et al., 2017; Jiskra et al., 2018; Fu et al., 2019; Sun et al., 2019). The seasonality in atmospheric oxidation chemistry, vegetation activity and residential coal combustion should be similar between urban and rural areas in China, as reflected by the seasonality in O3 (representing atmospheric oxidation chemistry), NDVI (representing vegetation activity) and CO (predominantly (40 %) originating from residential coal combustion) (Jiskra et al., 2018; Zheng et al., 2018), which showed summertime higher O3 concentrations and NDVI and lower CO concentrations at most urban sites (Fig. S5 in the Supplement). Therefore, the contrasting seasonal variations in TGM concentrations and isotopic compositions at most urban sites with respect to rural sites provided evidence that summertime enhanced emissions in these cities probably outbalanced the effect of seasonal variations in atmospheric oxidation chemistry, vegetation activity and residential coal combustion.

Traditionally, local and regional anthropogenic emissions were thought to dominate the TGM or GEM pollution in urban areas of China (Lin et al., 2010). A recent study showed quantitatively comparable coal combustion Hg emissions in China between winter and summer (Gao et al., 2019). Seasonal-resolution Hg emission inventories for other anthropogenic sources (e.g., production of cement, iron, steel, aluminum and non-ferrous metals) in China have not been established. Based on the monthly production data of these source materials, we estimated that there is no strong seasonality in total Hg emissions from these sources (Table S6 in the Supplement). Prevailing wind directions during the wintertime and summertime sampling campaigns were similar in Jinan, Lanzhou, Zhengzhou and Shanghai but were different in the remaining cities (Fig. S6 in the Supplement). Variations in the predominant wind direction would change the relationships between receptor and regional anthropogenic emissions, which could further influence the TGM levels and isotopic compositions in these cities. Given the similarity in wintertime and summertime prevailing wind directions in some cities and consistent summertime lower CO concentrations in most cities, it is postulated that the variations in local anthropogenic emissions and transport of regional anthropogenic emissions were not likely the main cause for the seasonal variations in TGM concentrations and isotopic compositions.

We found that the amplitudes of seasonal variations in TGM concentrations ((TGMsummer–TGMwinter) / TGMsummer) and Δ199Hg values (Δ199Hgsummer–Δ199Hgwinter) were both significantly positively correlated with latitude of the cities (ANOVA, R2 were 0.85 and 0.66 for TGM and Δ199Hg, respectively, p<0.01 for both; Fig. 5a and c), whereas the seasonal δ202Hg amplitudes (δ202Hgsummer–δ202Hgwinter) were significantly negatively correlated with latitude (ANOVA, R2=0.46, p<0.01; Fig. 5b). This indicates the seasonality in TGM concentrations and isotopic compositions were likely related to weather- and climate-dependent (e.g., solar radiation and air temperature) sources and/or atmospheric processes.

GEM emission fluxes from soil, building surfaces and pavement in urban areas are highly related to solar radiation and temperature and frequently peak in summer in the Northern Hemisphere (Gabriel et al., 2006; Eckley and Branfireun, 2008). Studies on the seasonal variations in GEM emissions from building surfaces and pavement are not available in Chinese urban areas but are expected to be similar to that of soil GEM emission (Gabriel et al., 2006). Therefore, using simulated seasonal soil GEM emission data is generally adequate to interpret the effect of surface GEM emission on the seasonal variations in TGM concentrations and isotopic compositions. As shown in Fig. 6a, a significant positive correlation was observed between the seasonal amplitudes of TGM concentration and simulated soil GEM emission flux ((fluxJuly–fluxJanuary) / fluxJuly), indicating enhanced surface GEM emission is responsible for the summertime increase of TGM concentrations at most urban sites. Negative seasonal TGM magnitudes were observed in Guiyang and Guangzhou in the low latitudes where there is a small summertime increase of soil GEM emission fluxes (Fig. 6a). We postulate that the effect of surface emission on the seasonal variations in TGM concentrations in Guiyang and Guangzhou was likely outbalanced by other factors, e.g., seasonal variations in atmospheric oxidization chemistry, vegetation activity and residential coal combustion.

Site-specific mean TGM δ202Hg and Δ199Hg values were calculated for summer and winter sampling campaigns separately and then values at all the sampling sites were correlated with their respective simulated soil GEM emission fluxes. A significant negative correlation was obtained between TGM δ202Hg and simulated soil emission (ANOVA, R2=0.43, p<0.01; Fig. 6b). As mentioned above, the isotopic compositions of GEM emitted from urban surfaces were characterized by highly negative δ202Hg values (mean = -1.57 ‰). Thus, high surface GEM emissions should shift TGM δ202Hg towards negative values. A weak positive correlation was observed between mean TGM Δ199Hg and simulated soil GEM emission fluxes (ANOVA, R2=0.21, p<0.05; Fig. 6b), suggesting that high surface GEM emissions led to a slightly positive shift of TGM Δ199Hg. Seasonal amplitudes of δ202Hg (or Δ199Hg) in the 10 cities were significantly negatively (or positively) correlated with seasonal amplitudes of simulated soil GEM emission flux (ANOVA, R2 of 0.54 or 0.63, p<0.01 for both; Fig. S7 in the Supplement), suggesting the dominant role of surface GEM emissions on the seasonal variations in TGM isotopic compositions.

It should be noted that indoor TGM also have highly negative δ202Hg (-1.56 ‰, n=1) and positive Δ199Hg (0.17 ‰, n=1) values (Jiskra et al., 2019a), but this source is not likely a dominant one contributing to the seasonal variations in TGM isotopic compositions. Indoor TGM concentrations in urban areas can be highly elevated mainly due to evaporation of GEM from Hg-containing products (e.g., spills of liquid mercury in thermometers, fluorescent light and Hg switches) in the absence of sunlight (Carpi and Chen, 2001; Baughman, 2006). This source is expected to yield Δ199Hg / Δ201Hg slopes of ∼1.6 in TGM isotopic compositions due to the nuclear volume effect (NVE) (Zheng and Hintelmann, 2010; Ghosh et al., 2013). As shown in Fig. S8 in the Supplement, a York bivariate linear regression between TGM Δ199Hg and Δ201Hg in the studied cites showed a slope of 1.01 ± 0.10 (1σ), which is consistent with that of soil GEM emissions (1.09 ± 0.06, 1σ; Fig. S7 in the Supplement) and that of photoreduction of Hg(II) to GEM (∼1.0) (Blum et al., 2014) but much lower than that predicted for indoor GEM sources, suggesting that seasonal variations in TGM isotopic compositions were unlikely dominated by indoor emission sources.

Hence, we can conclude that the seasonal variations in TGM concentrations and isotopic compositions in the 10 cities were likely controlled by surface emission sources. However, it is currently difficult to determine which of the surface emission sources (e.g., soil, pavement or building surfaces) was more important. As discussed earlier, GEM emitted from these sources were characterized by similar isotopic signatures and are difficult to distinguish. GEM emissions flux data from pavement and building surfaces in Chinese urban areas are very limited. A previous study in Toronto, Canada, and Austin, USA, reported that GEM emission fluxes from soils were on average 8 times higher than those from pavement and building surfaces (Eckley and Branfireun, 2008). This, together with the large fraction of cropland and sparsely vegetated soil area in the total urban land area (mean = 57 %; Table S1 in the Supplement), indicates soil emissions were likely more important than building surface and pavement emissions at a regional scale (e.g., the size of 1∘ × 1∘ surrounding the sampling sites). However, given that building surfaces and pavement emissions sources were in close proximity to the sampling sites (Fig. S2 in the Supplement), their contributions to atmospheric TGM budget may exceed those of soil emission sources locally. Therefore, further studies and approaches are needed to better constrain the contributions of local and regional land surface emissions to TGM variations at specific sites.