Interactive comment on “ Correlation slopes of GEM / CO , GEM / CO 2 , and GEM / CH 4 and estimated mercury emissions in China , South Asia , Indochinese Peninsula , and Central Asia derived from observations in northwest and southwest China ”

The authors present measurements of GEM concentrations made at three remote sites in China. Using concurrently measured CO, CO2, and CH4 mixing ratios they calculate GEM/CO, GEM/CO2, and GEM/CH4 concentration ratios for a number of pollution events. The authors compare these ratios with ratios reported by others for China and some other areas. The origin of the individual pollution events are attributed to four major source areas using backward trajectories. The transport from the source area


Introduction
Mercury (Hg) is a persistent pollutant in the environment and poses risks for human health risks, mainly through consumption of fish.Due to primary emissions and re-emissions of Hg from anthropogenic sources, the global atmospheric Hg budget has increased significantly since the Industrial Revolution (Mason et al., 1994).There are three major operationally defined Hg forms in the atmosphere, namely gaseous elemental mercury (GEM), gaseous oxidized mercury (GOM), and particulate-bound mercury (PBM).Knowledge on the anthropogenic and natural emissions of Hg into the atmosphere is important for a better understanding of Hg fate in the natural environment (Lindberg et al., 2007).Since the late 1980s, studies have been carried out to investigate the spatial and temporal characteristics of Hg emissions from anthropogenic (Nriagu, 1989;Pirrone et al., 1996;Pacyna et al., 2003;Streets et al., 2005;Pacyna et al., 2010;Pirrone et al., 2010;Streets et al., 2009) and natural sources (Lindberg et al., 1998;Gustin et al., 1999;Gustin et al., 2000;Gustin, 2003;Shetty et al., 2008).Improved emission factors for estimating Hg release from different source categories have significantly reduced the uncertainties (typically < 40 %) of recently reported anthropogenic emissions (Lindberg et al., 2007;Pacyna et al., 2010;Pirrone et al., 2010).The natural emissions (including primary natural emissions and re-emissions of historically deposited Hg), however, still have large uncertainties due to poor understanding of process mechanisms and a lack of reliable data on Hg 0 air-surface exchange (Gustin et al., 2005;Schroeder et al., 2005;Selin et al., 2007;Zhang et al., 2009).
Asia is the largest anthropogenic source region of Hg.It contributes approximately two-thirds of global total anthropogenic Hg emissions (Pacyna et al., 2010;Pirrone et al., 2010).Significant progresses have been made in the estimate of anthropogenic Hg emissions in China (Streets et al., 2005;Wu et al., 2006;Tian et al., 2011;Liang et al., 2013).The most recent estimate suggests that total anthropogenic Hg emissions in China increased to 1028 t in 2007 (Liang et al., 2013), nearly 2-fold higher than that in 1995 (Streets et al., 2005).In contrast, anthropogenic Hg emissions in other Asian countries (e.g., South Asia, Southeast Asia, and Central Asia) have received little attention.Such a lack of information limits the development of Hg emission inventories in a globally important source region.Due to the rapid economic development, anthropogenic Hg emissions in these regions are expected to considerably contribute to the regional Hg release (Pacyna et al., 2010).
Estimation of Hg emissions using observed concentrations of atmospheric GEM and other trace gases is a relatively novel approach for studying regional atmospheric Hg budgets.This method was first employed for estimating GEM emissions in the northeastern USA using the GEM / CO 2 correlation slopes (Lee et al., 2001).The approach was further improved and then applied for estimating Hg emissions in Asia, Europe, and South Africa (Jaffe et al., 2005;Slemr et al., 2006;Brunke et al., 2012).Such a measurement-based method complements the regional emission inventories estimated by means of conventional statistical approaches.It also yields an estimate of total Hg emissions from both anthropogenic and natural sources (Jaffe et al., 2005;Slemr et al., 2006).In the present study, the correlations slopes of GEM / CO, GEM / CO 2 , and GEM / CH 4 were investigated using the long-term atmospheric measurements at three remote stations in northwestern and southwestern China.The correlation slopes were classified into four source regions in Asia (mainland China, South Asia, the Indochinese Peninsula, and Central Asia) through trajectory analysis for estimating atmospheric Hg emissions.This aim of this work is to fill the knowledge gaps in our understanding on Asian Hg emissions.

Observational sites
In this study, observations were conducted at three remote sites in northwestern and southwestern China: Mt.Waliguan Baseline Observatory, Shangri-La station, and Mt.Ailao station (Fig. 1).The Mt. Waliguan observatory (WLG; 100.898 • E, 36.287• N; 3816 m a.s.l.) is one of the World Meteorological Organization's (WMO) Global Atmospheric Watch (GAW) Baseline Stations, and is situated at the summit of Mt.Waliguan at the edge of northeastern part of the Qinghai-Xizang (Tibet) Plateau.WLG is relatively isolated from industrial point sources and populated regions.The surrounding areas of WLG are naturally preserved arid/semiarid lands and scattered grasslands and there is no local Hg source around the station.Most of the Chinese industrial and populated regions which may be related to anthropogenic Hg emissions are situated to the east of WLG.Due to the influ-ence of the Qinghai-Xizang (Tibet) Plateau; the predominate wind directions are from the west to southwest sector in cold seasons and the east sector in warm seasons.As shown in Fig. 1, the three provinces of Qinghai, Xinjiang, and Xizang, have fairly low anthropogenic emissions of Hg, CO, CO 2 , and CH 4 relative to eastern China, Central Asia, and South Asia (Wu et al., 2006;Zhao et al., 2012b;Zhao et al., 2012a;Zhang et al., 2014a;Kurokawa et al., 2013).Therefore, anthropogenic emissions in these three Chinese provinces are expected to have a minimal effect on the westerly and southwesterly airflows in cold seasons, which in turn largely reflects the feature of long-range atmospheric transport of air pollutants from Central Asia and South Asia to WLG.
The Shangri-La station (XGL; 99.733 • E, 28.017 • N; 3580 m a.s.l.) is located in the Hengduan Mountains area in the southeastern Tibetan Plateau, southwestern China (Fig. 1).The shortest distances from XGL to South Asia and the Indochinese Peninsula are 260 and 100 km, respectively.XGL is surrounded by naturally preserved forest and mountainous areas.There are no large point sources within 100 km of the station with the exception of the city of Shangri-La, which is situated 30 km to the south of the station, with a population of about 140 000, and may be related to anthropogenic emissions of Hg and other air pollutants.Areas to the west and south of the station are well-preserved mountainous forest and have no significant anthropogenic sources.The long-range transport of air masses from South Asia and the Indochinese Peninsula to XGL is not likely impacted by these areas.
The Mt. Ailao station (MAL; 100.017 • E, 24.533 • N; 2450 m a.s.l.) is located at a summit of the north side of Ailao Mountain National Nature Reserve in central Yunnan province, southwestern China.The reserve stretches more than 130 km from south to north with a maximum width of approximately 20 km.More than 85 % of the reserve is covered by preserved forest.MAL is isolated from industrial sources and populated regions in China.Kunming, one of the largest cities in southwestern China, is located 180 km to the northeast of the station.The site is approximately 200 and 600 km away from the Indochinese Peninsula and South Asia, respectively.Most of the Chinese anthropogenic sources of Hg and other air pollutants are located to the north and east of the station, whereas anthropogenic emissions in southern and western Yunnan province are fairly low (Wu et al., 2006;Zhao et al., 2012b;Zhao et al., 2012a;Zhang et al., 2014a;Kurokawa et al., 2013).The wind system at the station is dominated by the India summer monsoon (ISM) in warm seasons and the westerlies surrounding the Tibetan Plateau in cold seasons.The ISM can carry air pollutants from the Indochinese Peninsula and southern China, while the westerlies carry air pollutants from South Asia.

Measurements of GEM, CO, CO 2 , and CH 4
Continuous measurements of atmospheric GEM at WLG, XGL, and MAL were conducted using an automated Hg vapor analyzer (Tekran 2537A/B).This analyzer has been used extensively for atmospheric total gaseous mercury (TGM) measurements worldwide (Ebinghaus et al., 1999;Munthe et al., 2001;Fu et al., 2012a).It combines the pre-concentration of TGM onto gold traps with thermal desorption and cold vapor atomic fluorescence spectrometry detection of GEM.The analyzer has two gold cartridges working in parallel.While one cartridge collects TGM, the other one performs analysis of the collected TGM.The function of the cartridges is then reversed, allowing continuous sampling of ambient air.The analyzer was set up in a temperature-controlled laboratory (15-25 • C).Ambient air was introduced to the inlet of the analyzer through a 25 ft heated Teflon tube (50 • C).Air particulate matter was removed by using two 45 mm diameter Teflon filters (pore size 0.2 µm), which were installed at the inlets of the sampling Teflon tube and analyzer, respectively.The analyzer was programmed to measure atmospheric TGM at the time resolution of 5 min at XGL and MAL and 10 min at WLG with a volumetric sampling flow rate of ∼ 1.5 L min −1 .The data quality of the analyzer was controlled by periodic (every 25 h) automatic permeation source injections, and the internal permeation source was calibrated every 3-6 months (Fu et al., 2012a).Atmospheric TGM in general consists of GEM and GOM.In most cases in these studies, GOM constitutes a small portion of TGM (< 1 %; Fu et al., 2012a), and a large fraction of GOM was expected to be captured by the sampling Teflon tube and soda lime trap installed at the inlet of the Tekran 2537A/B analyzer (Fu et al., 2010a;Fu et al., 2012b).Hence, the atmospheric TGM measured herein is referred to as GEM throughout the paper.Concentrations of GEM are expressed in ng m −3 Standard Temperature and Pressure (STP) with a standard temperature of 273.16K and pressure of 1013 hPa.
Atmospheric CO 2 at WLG was measured using a LI-COR 6251 non-dispersive infrared (NDIR) analyzer, and CH 4 and CO were measured using a G1301 (Picarro, USA) and a G1302 (Picarro, USA) cavity ring-down spectroscopy system (CRDS), respectively.Atmospheric CO 2 and CH 4 at XGL were measured using the G1301 CRDS, and CO was measured by the G1302 CRDS.Detailed information regarding the schematic of the analytical systems, air collections, calibrations, and data processing has been addressed in previous studies (Zhou et al., 2003;Fang et al., 2013).The analytical precisions for the atmospheric CO 2 , CH 4 , and CO measurements were approximately 0.07 µmol mol −1 (ppm), 1.5 nmol mol −1 (ppb), and 2.0 nmol mol −1 (ppb), respectively.Atmospheric CO concentrations at MAL were measured using a non-dispersive infrared instrument (Thermo Environmental Instruments model 48C) (Jaffe et al., 2005) between the atmosphere and the forest canopy, atmospheric CO 2 concentrations at XGL exhibited strong diurnal variations.This had a significant impact on the correlation analysis between GEM and CO 2 .In light of this, we did not study the correlation of GEM to CO 2 at XGL.

Method of correlation analysis
Correlation analysis between atmospheric compounds is a novel tool for studying regional emissions strength of atmospheric pollutants.It has been used for estimating emissions of many atmospheric pollutants with data in good agreement with established emissions inventories (Yokouchi et al., 2006;Worthy et al., 2009;Tohjima et al., 2014).This method was developed and first utilized for estimating Hg emissions in Asia using the correlation of GEM to CO during Asian outflow events (Jaffe et al., 2005).Subsequently, correlation slopes between GEM and other trace gases such as CH 4 , CO 2 , and halocarbons were also calculated (Slemr et al., 2006;Brunke et al., 2012).These methods are based on assumptions of no chemical and physical losses of air pollutants, as well as constant emission ratios and constant background of air pollutants during atmospheric transport events (Jaffe et al., 2005).In this study, correlations between atmospheric GEM and CO, CO 2 , and CH 4 were utilized to estimate GEM emissions from mainland China, South Asia, the Indochinese Peninsula, and Central Asia on the basis of continuous measurements of atmospheric GEM, CO, CO 2 , and CH 4 at WLG, XGL, and MAL.The three stations are located in remote areas of northwestern and southwestern China and have constant backgrounds of atmospheric pollutants (Fu et al., 2012a;Zhang et al., 2014b).Also, the transport time (typically less than 5 days) of air masses from the source regions to the stations is much shorter than the atmospheric lifetimes of GEM, CO, CO 2 , and CH 4 .The multiple correlation relationships help constrain the estimated GEM emissions.
Correlation analysis was conducted by computing the Pearson correlation (orthogonal least-squares correlation) between GEM concentrations and CO, CO 2 , and CH 4 concentrations independently during pollution events when air masses originated from or passed over a source region consistently.These events had GEM concentrations enhanced by at least 0.5 ng m −3 and lasted for 8-24 h.Correlation slopes were selected when linear positive correlation was significant (p>0.01) with a correlation coefficient (r 2 ) > 0.5 (signifi- cant correlations with negative slopes were excluded).This criterion was to ensure the method assumptions are valid (Jaffe et al., 2005).Figure 2 shows the time series of atmospheric GEM, CO, CO 2 , and CH 4 concentrations during the transport events from 30 January to 2 March in 2009.The temporal variations of GEM were not consistently correlated with those of the three air pollutants because these events were possible impacted by different sources that led to different relative emission strengths of air pollutants.Therefore, the correlation slopes of GEM / CO, GEM / CO 2 , and GEM / CH 4 were calculated individually in this study.
In addition, CH 4 / CO, CH 4 / CO 2 , and CO / CO 2 correlation slopes were calculated during pollution events with CH 4 and CO concentrations elevated by at least 10 and 20 ppb, respectively, on the basis of criterion mentioned for GEM / CO, GEM / CO 2 , and GEM / CH 4 correlation slopes.These correlation slopes are useful for constraining CH 4 , CO, and CO 2 emissions in the study regions.

Air mass trajectory calculation
To establish the relationships between the observed correlations slopes and the source regions in Asia, we calculated 5day backward trajectories every 2 h at each of the stations using the TrajStat Geographical Information System software and gridded meteorological data (Global Data Assimilation System, GDAS1) from the US National Oceanic and Atmospheric Administration (NOAA).The global gridded meteorological data have a horizontal resolution of 1 • (360 × 180 grid cells) with 23 vertical levels from 1000 to 20 hPa (Wang et al., 2009).These trajectories ended at a height of 500 m above ground at WLG, XGL, and MAL.The trajectory endpoints in each event were averaged to yield the transport pathway.The source area identified by the trajectory analysis was weighted by the correlation slope observed for the stations during the event.
It should be noted that the endpoints of the backward trajectories can occasionally pass over multiple regions.In this case, we attributed the correlation slopes to the most important source regions that the air masses traveled through.For example, most air masses that originated from and/or passed over South Asia and Central Asia and ended at WLG also passed over the Chinese provinces of Xinjiang, Xizang, and Qinghai, which have fairly low emissions of atmospheric GEM, CO, CO 2 , and CH 4 (Wu et al., 2006;Kurokawa et al., 2013).It is therefore assumed that these air masses carried the emission signals from Central and South Asia.On the other hand, eastern and central China is an important source region of atmospheric GEM, CO, CO 2 , and CH 4 , and therefore the air masses passing over eastern and central China were assumed to carry the emission signals in China.For XGL and MAL, the areas to the west and south of the stations in Yunnan province, southwestern China have fairly low emissions of atmospheric GEM, CO, CO 2 , and CH 4 (Wu et al., 2006;Kurokawa et al., 2013).The air masses that passed over South Asia and the Indochinese Peninsula were assumed to carry the emission signals from the respective region.

Atmospheric GEM concentrations at WLG, XGL, and MAL and potential source regions
Averaged atmospheric GEM concentrations during the study period were 2.05 ± 0.96 ng m −3 (hourly means ranging from 0.40 to 14.58 ng m −3 , October 2007 to September 2009) for WLG, 2.52 ± 0.70 ng m −3 (hourly means ranging from 1.35 to 5.98 ng m −3 , from July to October 2010) for XGL, and 2.05 ± 0.67 ng m −3 (hourly means ranging from 0.89 to 6.26 ng m −3 , from September 2011 to March 2013) for MAL.The levels of atmospheric GEM at the three stations were relatively lower compared to those observed in North America and Europe (1.3-2.0 ng m −3 ; Sprovieri et al., 2010;Lan et al., 2012;Cole et al., 2013;Munthe et al., 2003).A previous study by Fu et al. (2012a) at WLG suggested that long-range atmospheric transport of GEM from industrial and urbanized areas in northwestern China and northwestern India contributed significantly to the elevated GEM at WLG.
For GEM at XGL, potential sources areas included northern India, Myanmar, western Sichuan province, and western Yunnan province (Zhang et al., 2014b).The potential source areas varied with monsoon at MAL.During the ISM

Observed correlation slopes for the studied source regions
Using backward trajectory analysis, we divided GEM / CO slopes into four regions, GEM / CO 2 into three regions, and GEM / CH 4 into four regions (Table 1).Histograms of GEM / CO, GEM / CO 2 , and GEM / CH 4 slopes for the identified source regions are displayed in Fig. 3.Most of the correlation slopes followed a lognormal distribution (Table 1).Hence, geometric means of the correlation slopes are used throughout the paper.The geometric mean correlation slopes of GEM / CO for mainland China, South Asia, the Indochinese Peninsula, and Central Asia were 7.3 ± 4.3 (1 SD, n = 37), 7.8 ± 6.4 (1 SD, n = 40), 7.8 ± 5.0 (1 SD, n = 34), and 13.4 ± 9.5 pg m −3 ppb −1 (1 SD, n = 6), respectively.The observed correlation slopes for mainland China were associated  with air masses originating from northwestern, southwestern, central, and southern China (Fig. 4).The trajectories were simulated for a period of 5 days and were therefore expected to pass over most of mainland China because of the length of the trajectories.As a result, these observed correlation slopes of GEM / CO for northwestern, southwestern, central, and southern China were likely representative of the emissions from a majority of areal coverage of mainland China.The correlation slopes of GEM / CO observed for South Asia, the Indochinese Peninsula, and Central Asia are also representative of these regions as the air masses pass over a majority of these regions (Fig. 4).GEM / CO correlation slopes were comparable among mainland China, South Asia, and the Indochinese Peninsula (means range from 7.3 to 7.8 pg m −3 ppb −1 ), but nearly 2-fold lower than the mean for Central Asia (mean = 13.4 ± 9.5 pg m −3 ppb −1 ).This trend is consistent with the anthropogenic emission ratios of GEM to CO in different regions of Asia.Based on the published anthropogenic GEM and CO emissions inventories in Asia (Kurokawa et al., 2013;AMAP/UNEP, 2013;Wu et al., 2006), we calculated the anthropogenic GEM / CO emission ratio to be 7.2 pg m −3 ppb −1 (5.8 × 10 −6 g g −1 ) for Central Asia, which is significantly higher than those ratios for mainland China (2.7 pg m −3 ppb −1 ; 2.2 × 10 −6 g g −1 ), South Asia (1.6 pg m −3 ppb −1 ; 1.3 × 10 −6 g g −1 ), and the Indochinese Peninsula (1.5 pg m −3 ppb −1 ; 1.2 × 10 −6 g g −1 ) (Table 2).Although correlation slopes of GEM / CO were also likely influenced by secondary emissions of GEM (Jaffe et al., 2005;Slemr et al., 2006), the higher anthro-   (Zhao et al., 2012b), c (Zhao et al., 2012a), d (Zhang and Chen, 2010), e (Kurokawa et al., 2013).pogenic GEM / CO emission ratio in Central Asia partially explains the elevated correlation slopes of GEM / CO in the region.The GEM / CO correlation slopes for mainland China were slightly higher than those for Chinese outflows observed at Hedo Station, Okinawa, Japan; Mt.Bachelor Observatory (MBO), western USA; and Seoul, Korea (4.6-7.4 pg m −3 ppb −1 ), as well as from coastal flight observations (Jaffe et al., 2005;Weiss-Penzias et al., 2007;Choi et al., 2009;Pan et al., 2006;Friedli et al., 2004), but slightly lower than values observed in the air masses that originated from and/or passed over eastern China (8.0 and 11.4 pg m −3 ppb −1 ) (Friedli et al., 2004;Sheu et al., 2010) and those in the upper troposphere observed during the flights from Frankfurt, Germany, to Guangzhou, southern China (8.2-12.8pg m −3 ppb −1 ) (Slemr et al., 2009;Slemr et al., 2014).The difference between the present study and literature values may reflect a regional emission difference.The correlation slopes calculated from the observations in mainland China were associated with air masses originated from and passed over northwestern, southwestern, central, and southern China (Fig. 4), whereas those estimated in previous studies were associated with the air masses in eastern China.Furthermore, there may be impacts from recent changes in atmospheric sources of GEM, including the decreasing contributions of GEM emissions from domestic coal, agricultural residual, and forest burning emissions to the total anthropogenic emissions in mainland China during the past decade (Wu et al., 2006;Liang et al., 2013).These sources were reported with relatively lower GEM / CO emission ratios compared to other industrial sources (Weiss-Penzias et al., 2007;Zhang et al., 2013).Apart from the study of Sheu et al. (2010), in which a GEM / CO correlation slope (5.0 pg m −3 ppb −1 ) for the outflow from the Indochinese Peninsula was reported, there have been few studies regarding the correlation slopes of GEM / CO in South Asia, the Indochinese Peninsula, and Central Asia.
The geometric means of GEM / CO 2 correlation slopes for mainland China, South Asia, and Central Asia were 248 ± 119 (1 SD, n = 25), 270 ± 164 (1 SD, n = 21), and 315 ± 289 pg m −3 ppm −1 (1 SD, n = 13), respectively.The GEM / CO 2 correlation slopes calculated from the observations in mainland China were associated with air masses originating from northwestern and southwestern China, as well as from central China (Fig. 5).The GEM / CO 2 correlation slopes associated with trajectories transported from South Asia predominantly came from Pakistan and northwestern India, covering a large area of Central Asia (Fig. 5).The values of GEM / CO 2 correlation slopes also vary with regions, with the greatest geometric mean GEM / CO 2 for the Central Asia.The spatial pattern of GEM / CO 2 slopes appeared to be consistent with the anthropogenic emission ratios of GEM to CO 2 in Asia.With the anthropogenic emissions of GEM and CO 2 taken into account (Wu et al., 2006;Kurokawa et al., 2013; AMAP/UNEP, 2013), GEM / CO 2 emission ratios of anthropogenic sources were on the order of 5.1 × 10 −8 t t −1 in Central Asia and 4.2 × 10 −8 t t −1 in mainland China, and 3.9 × 10 −8 t t −1 in South Asia.
The geometric means of the correlation slopes of GEM / CH 4 for mainland China, South Asia, the Indochinese Peninsula, and Central Asia were 33.3 ± 30.4 (1 SD, n = 41), 27.4 ± 31.0 (1 SD, n = 4), 23.5 ± 15.3 (1 SD, n = 6), and 20.5 ± 10.0 pg m −3 ppb −1 (1 SD, n = 6), respectively.Fortyone GEM / CH 4 ratios were calculated in mainland China (26 at the WLG site and 15 at the XGL site).The correlation slopes of GEM / CH 4 at WLG were associated with the air masses from northwestern China and those at XGL were associated with the air masses from Yunnan province, southwestern China (Fig. 6).The events of air transport from South Asia, the Indochinese Peninsula, and Central Asia (4-6 slopes for each region) were relatively fewer.The air masses related to the slopes for South Asia and the Indochinese Peninsula were from northwestern India and Myanmar, respectively.The GEM / CH 4 ratios calculated from the observations in the air masses from different regions in mainland China varied significantly (Fig. 6).The GEM / CH 4 ratios associated with the air masses from northwest-ern China fell within the range of 14.6-208 pg m −3 ppb −1 (geometric mean = 49 pg m −3 ppb −1 ), significantly higher than those from southwestern China (ranging from 8 to 69 pg m −3 ppb −1 , geometric mean = 20 pg m −3 ppb −1 ).The lower GEM / CH 4 values estimated from the air transport from southwestern China are likely due to the CH 4 emissions from rice paddies and natural wetlands (Zhang and Chen, 2010;Chen et al., 2013;Zhang et al., 2014a).The anthropogenic release of GEM into the atmosphere in the region is relatively lower (Fu et al., 2012c).

Implications of atmospheric Hg emission sources in Asia
The correlation slopes of GEM / CO, GEM / CO 2 , and GEM / CH 4 were similar in Asian regions.This indicates that the sources of atmospheric Hg were more or less similar among the four studied regions.The GEM / CO, GEM / CO 2 , and GEM / CH 4 correlation slopes in Asian were higher than those observed in other regions.For example, the GEM / CO ratios in Europe, South Africa, and North America were in the range of 0.3-5.0pg m −3 ppb −1 (Jaffe et al., 2005;Slemr et al., 2006;Brunke et al., 2012).For GEM / CO 2 ratios, previous studies reported a mean of 184 pg m −3 ppm −1 for the northeastern USA and 63 pg m −3 ppm −1 for South Africa (Lee et al., 2001;Brunke et al., 2012).The mean GEM / CH 4 ratios in Europe and South Africa were 3.9 and 3.6 pg m −3 ppb −1 , respectively (Slemr et al., 2006;Brunke et al., 2012), approximately 1 order of magnitude lower than those in Asia.Non-ferrous metal smelting (zinc, lead, and gold production), coal combustion, cement production, and mercury production are the four largest source categories of anthropogenic GEM emissions in mainland China.Emission factors (EFs) of CO and CO 2 from anthropogenic sources have been investigated extensively.In this study, the EFs of CO and CO 2 summarized from Zhao et al. (2012a) and Zhao et al. (2012b) were adopted for calculating GEM / CO and GEM / CO 2 emission ratios for anthropogenic sources.EFs of GEM from coal combustion, non-ferrous smelting, cement, primary mercury production, and large-scale gold production have also be reported (Streets et al., 2005;Wang et al., 2010;Li et al., 2010;Pacyna et al., 2010).The emission ratios of GEM / CH 4 were not estimated due to the fact that GEM and CH 4 do not have common emission sources.The estimated emission ratio of GEM / CO is 149 pg m −3 ppb −1 for zinc smelting, 120 pg m −3 ppb −1 for lead smelting, 0.6 pg m −3 ppb −1 for coal combustion, 0.3 pg m −3 ppb −1 for cement production, 6.7 × 10 3 pg m −3 ppb −1 for primary mercury production, and 870 × 10 3 pg m −3 ppb −1 for large-scale gold production.The estimated emission ratio of GEM / CO 2 is 48 × 10 3 pg m −3 ppm −1 for zinc smelting, 131 × 10 3 pg m −3 ppm −1 for lead smelting, 10 pg m −3 ppm −1 for coal combustion, and 36 pg m −3 ppm −1 for cement production, 190 × 10 3 pg m −3 ppm −1 for primary mercury production, and 240 × 10 5 pg m −3 ppm −1 for large-scale gold production.It should be pointed out that artisanal and smallscale gold and mercury production is also important sources of atmospheric GEM in Asia (Pacyna et al., 2010;Muntean et al., 2014).These two sources are generally equipped with poor Hg-control devices and are expected to produce much larger emission ratios of GEM / CO and GEM / CO 2 (Li et al., 2009;Muntean et al., 2014).Biomass burning (forest and grassland fires, crop residual burning, and crop residues and wood combustion) is also an important atmospheric GEM source in mainland China (Huang et al., 2011).The observed GEM / CO and GEM / CO 2 emission ratios for biomass burning were 0.6-2.1 and 109 pg m −3 ppm −1 , respectively (Brunke et al., 2001;Friedli et al., 2003;Weiss-Penzias et al., 2007;Ebinghaus et al., 2007).Given that non-ferrous smelting and mercury production are important sources of atmospheric GEM in Asia and have relatively higher GEM / CO and GEM / CO 2 emission ratios, the elevated GEM / CO and GEM / CO 2 correlation slopes in Asia are likely to have resulted from these emission sources.None of the GEM / CO and GEM / CO 2 emission ratios from anthropogenic sources agree consistently with the observed correlation slopes, indicating that the observed correlation slopes of GEM / CO and GEM / CO 2 were likely influenced by multiple sources, including release from natural surfaces.Anthropogenic emission alone is not able to fully explain the observed correlation slopes.Based on the annual anthropogenic emission inventories of GEM, CO, CO 2 , and CH 4 , the emission ratios of GEM / CO, GEM / CO 2 , and GEM / CH 4 were calculated; results are shown in Table 2.The anthropogenic emission ratios were all signifi-cantly lower than the correlation slopes of observed concentrations.The discrepancy was partially attributed to the fact that observed correlation slopes were not uniformly distributed within different regions and seasons, which may be not adequate to represent the annual and overall characteristics of the emission ratios in the studied regions.Unfortunately, this reason cannot be evaluated further due to the lack of seasonal assessments of the GEM, CO, CO 2 , and CH 4 emissions.In addition, it is speculated that contributions from soil emission of GEM may play a crucial role.Soil emissions are an important source of atmospheric GEM (Pirrone et al., 2010).Due to the lack of soil GEM flux measurements from South Asia, the Indochinese Peninsula, and Central Asia, the measurements in China were applied for the analysis.The measured soil GEM fluxes in southwestern and southern China fell within the ranges of 19.2-132 (mean = 49 ng m −2 h −1 ) and 18.2-114 ng m −2 h −1 (mean = 43 ng m −2 h −1 ), respectively (Feng et al., 2005;Fu et al., 2008Fu et al., , 2012c).These values are significantly higher than those in Europe and North America (Zhang et al., 2001;Ericksen et al., 2006;Schroeder et al., 2005).Assuming the soil emissions of CO 2 in mainland China are comparable to those in Europe and North America, the elevated GEM emission fluxes from soil in China can lead to the GEM / CO and GEM / CO 2 correlation slopes in mainland China.Using the published CO 2 emission fluxes from subtropical arable soil (Lou et al., 2004), we calculated the soil GEM / CO 2 emission ratios to be 148-1070 pg m −3 ppm −1 (mean = 370 pg m −3 ppm −1 ).Given that soil does not release significant CO into the atmosphere (EC-JRC/PBL, 2011), soil emissions are expected to produce extremely high GEM / CO emission ratios.Rice paddies are sources of both GEM and CH 4 .However, previous studies suggested that GEM emission fluxes from rice paddies were much lower compared to those of bare soils, in the range of 1.4-23.8ng m −2 h −1 (Zhu et al., 2011;Fu et al., 2012c).The mean CH 4 emission flux in China has been recorded as high as 11.4 mg m −2 h −1 (Chen et al., 2013).This yields average GEM / CH 4 emission ratios of 0.1-1.5 pg m −3 ppb −1 from rice paddies.The low GEM / CH 4 emission ratios from rice paddies were opposite to our observations, indicating that they is not the cause of elevated GEM / CH 4 slopes in China.However, bare soils are not expected to release CH 4 and should produce extremely high GEM / CH 4 emission ratios.Given the larger areas and higher GEM fluxes of bare soils in China, elevated GEM / CH 4 correlation slopes in China are probably caused by dry soil GEM emissions.

Estimates of GEM emissions
GEM emissions in mainland China, South Asia, the Indochinese Peninsula, and Central Asia were calculated using the GEM / CO, GEM / CO 2 , and GEM / CH 4 correlation slopes obtained in the present study as well as emissions of CO, CO 2 , and CH 4 in Asian countries.Emissions of CO, CO 2 , www.atmos-chem-phys.net/15/1013/2015/and CH 4 in South Asia, the Indochinese Peninsula, and Central Asia were adopted from the study by Kurokawa et al. (2013), which in most cases are consistent with those reported by EDGAR 4.2 (EC-JRC/PBL, 2011).Emissions of CO, CO 2 , and CH 4 in mainland China were adopted from Chinese studies (Table 3).The emissions of CO and CO 2 in these studies agree with others reported in the literature (EC-JRC/PBL, 2011; Kurokawa et al., 2013;Liu et al., 2013;Tohjima et al., 2014).The CH 4 emission used for mainland China in this study (39.6 × 10 6 t yr −1 ) is significantly lower than those emissions reported by Kurokawa et al. (2013) and EDGAR 4.2 (EC-JRC/PBL, 2011) (73.2-76.0× 10 6 t yr −1 ) but similar to that predicted from CH 4 / CO 2 correlations at Hateruma Island (46.0 × 10 6 t yr −1 ) (Tohjima et al., 2014).
The Chinese studies utilized optimized emission factors for many sources (e.g., coal mining, rice cultivation, enteric fermentation) and are expected to give a better prediction of CH 4 emissions in China (Cheng et al., 2011;Chen et al., 2013).
Annual GEM emissions estimated from GEM / CO correlation slopes were 1071, 470, 125, and 54 t yr −1 for mainland China, South Asia, the Indochinese Peninsula, and Central Asia, respectively.The estimated GEM emissions from GEM / CO 2 correlation slopes are similar to those derived from GEM / CO correlation slopes, with annual GEM emissions of 1187, 340, and 90 t yr −1 for mainland China, South Asia, and Central Asia (no correlation slopes were observed for the Indochinese Peninsula).GEM emissions estimated from GEM / CH 4 correlation slopes were substantially higher than those derived from GEM / CO and GEM / CO 2 correlation slopes (Table 3).For example, the estimated GEM emission in China based on GEM / CH 4 ratios reached 1846 t yr −1 , 55-72 % higher than those estimated from GEM / CO and GEM / CO 2 ratios.Similarly, the estimated GEM emissions in South Asia, the Indochinese Peninsula, and Central Asia from GEM / CH 4 ratios were 1.2-3.9times greater than those estimated from GEM / CO and GEM / CO 2 ratios.
The discrepancy in GEM emissions might be due to the following reasons.First, it was reported that CH 4 emissions in China and other Asian counties have larger uncertainties compared to CO and CO 2 emissions (Olivier et al., 2001;Kurokawa et al., 2013).Recent Chinese studies have suggested that the CH 4 emissions in China reported by previous studies were overestimated by a factor of ∼ 2 (Zhang and Chen, 2010;Cheng et al., 2011;Chen et al., 2013;Kurokawa et al., 2013).This may also be the case for South Asia, the Indochinese Peninsula, and Central Asia.As shown in Tables 1 and 2, the observed CH 4 / CO and CH 4 / CO 2 correlation slopes for the studied regions were significantly higher than the emission ratios calculated on basis of published inventories, while CO / CO 2 correlation slopes were consistent with the emission ratios.This implies that previously reported CH 4 emissions in the studied regions were likely overestimated, which may partially explain the overestimated GEM emissions derived from GEM / CH 4 correlation slopes and CH 4 emissions.Secondly, we do not obtain substantial correlation slopes of GEM / CH 4 , which might be not representative for the studied regions.Using mainland China as the example, 26 out of 41 slopes were observed at WLG.The slopes were related to air masses that originated from and/or passed over northwestern China, which yielded a mean GEM / CH 4 correlation slope of 49 ± 30.0 pg m −3 ppb −1 , significantly higher than that of the slopes from southwestern China (20.0 ± 19.1 pg m −3 ppb −1 ).The slopes associated with air masses from northwestern China were expected to be predominantly influenced by emissions of GEM and CH 4 its proximity to WLG.Previous studies have suggested that the anthropogenic emission ratios of GEM / CH 4 in northwestern China were relatively higher than the values from other Chinese regions (Wu et al., 2006;Zhang et al., 2014a).Therefore, the large fraction of slopes obtained from northwestern China was also responsible for overestimates of GEM emissions in the present study.Hence, it is speculated that GEM / CO and GEM / CO 2 correlation slopes may better depict the GEM emissions in Asia than GEM / CH 4 correlation slopes in the present study.
The estimated GEM emissions in mainland China, South Asia, the Indochinese Peninsula, and Central Asia using GEM / CO and GEM / CO 2 ratios agree reasonably with the results of previous studies (Jaffe et al., 2005;Weiss-Penzias et al., 2007), but are consistently greater than the reported anthropogenic GEM emissions (Table 3).The estimated GEM emissions in China are about 3-4-fold higher than the anthropogenic emission for 2003-2010 (Wu et al., 2006;AMAP/UNEP, 2013;Muntean et al., 2014), and those in South Asia, the Indochinese Peninsula, and Central Asia are 2-5-fold higher than the anthropogenic emissions for 2010 (AMAP/UNEP, 2013).It is hypothesized that underestimates of anthropogenic GEM emissions, contributions of re-emission and natural emissions, uncertainties in the fraction of Hg species in the inventory, and conversion of Hg species during long-range transport are reasons for the discrepancy (Jaffe et al., 2005;Slemr et al., 2006).A recent study showed that the total anthropogenic Hg emissions in China increased to 1028 t in 2007, which is about 50 % higher than that in 2003 and corresponds to a mean annual rate of increase of 10.6 % (Wu et al., 2006;Liang et al., 2013).If this rate of increase is applied to the estimate of anthropogenic GEM emissions in 2003 (Wu et al., 2006), anthropogenic GEM emission in China is expected to be 800 t for 2010.This value is significantly higher than the estimate in 2003 as well as the value from the UNEP report for 2010 (430 t) (Wu et al., 2006; AMAP/UNEP, 2013).There have been few studies on anthropogenic GEM emissions in South Asia, the Indochinese Peninsula, and Central Asia.A previous study suggested that total Hg emission in India was about 253 t in 2004 (Mukherjee et al., 2009).Assuming GEM accounting for 64 % of total Hg emissions in India (Pacyna et al., 2003), GEM emission in India for 2004 was estimated to be 162 t, ∼ 2 times greater compared to the estimate of 96 t in South Asia (including India and other South Asian countries) for 2010 by the UNEP report (AMAP/UNEP, 2013).Given the increasing energy consumption recently, an increase in GEM emissions in South Asia is expected.This indicates that the UNEP report for 2010 may have underestimated GEM emissions in South Asia significantly.
Emission and re-emission of GEM from natural sources were regarded as an important cause of the discrepancy between estimated GEM emissions using correlation slopes of atmospheric pollutants and anthropogenic emission inventories (Jaffe et al., 2005;Slemr et al., 2006).Figure 7 shows the statistical summary of GEM exchange fluxes between different landscapes and the atmosphere in the warm season (from May to October) in mainland China.The mean GEM flux from dry farmland, rice paddies, grassland, forest soil, lake and river waters, and urban soil in warm seasons were 33.6 ± 34.6, 17.4 ± 15.9, 11.4 ± 11.1, 8.8 ± 6.4, 6.1 ± 4.4, and 35.3 ± 43.1 ng m −2 h −1 , respectively.These are significantly higher compared to those observed from Europe and North America (Poissant and Casimir, 1998;Boudala et al., 2000;Schroeder et al., 2005;Ericksen et al., 2006;Kuiken et al., 2008;Choi and Holsen, 2009).GEM fluxes from different landscapes in cold seasons (from November to April) were relatively limited.Several studies found that GEM References: Wang et al. (2003, 2006), Feng et al. (2004, 2005), Fang et al. (2004), Fu et al. (2007, 2008, 2010b, 2012c, 2013), Zhu et al. (2011Zhu et al. ( , 2013)), Ma et al. (2013), andLiu et al. (2014) fluxes from dry farmland, forest soil, and lake waters were about 2.5-40 times (mean = 6.5, n = 18) lower than those in warm seasons (Wang et al., 2003;Ma et al., 2013;Fu et al., 2010b;Fu et al., 2013).Given the different landscapes and seasonal patterns of GEM fluxes in mainland China, we estimate the annual natural GEM emissions to be 528 t in China.This value is close to the estimate made by Shetty et al. (2008) but highlights the GEM emissions from dry farmland and grassland.There is no information regarding GEM fluxes from different landscapes in South Asia, the Indochinese Peninsula, and Central Asia.Here we assume that the natural GEM fluxes from landscapes in these areas are similar to those in China and the annual GEM emissions from South Asia, the Indochinese Peninsula, and Central Asia could be roughly estimated to be 240, 113, and 220 t, respectively.
Uncertainties and limitations related to the correlation slope method may also be important for the discrepancy between estimated GEM emissions and anthropogenic emission inventories.These uncertainties and limitations may include the uncertainties of CO, CO 2 , and CH 4 emissions as well as the limitations of observed correlation slopes.The uncertainties for CO 2 and CO emissions in China in Table 3 were estimated to be about 11 and 45 %, respectively (Zhao et al., 2012a, b).Uncertainties for Chinese CH 4 emissions were not calculated by Zhang and Chen (2010) but are expected to have much larger values (Kurokawa et al., 2013).The uncertainties for CO 2 , CO, and CH 4 emissions in South Asia, the Indochinese Peninsula, and Central Asia in Table 3 were estimated to be 44-49, 114-131, and 154-204 (Kurokawa et al., 2013).The limitations related to the correlation slopes were mainly caused by the fact that some of the emissions sources and pollution control devices of GEM, CO 2 , CO, and CH 4 are different, and this is a particular issue for GEM / CH 4 correlation slopes.It possibly resulted in temporal and spatial variations of emission ratios between GEM and CO 2 , CO, and CH 4 .As shown in Sects.3.2 and 3.3, GEM / CO and GEM / CH 4 correlation slopes for mainland China observed at WLG (mainly related to air masses from northern and northwestern China) were 66 and 145 %, respectively, higher than that observed at XGL and MAL (mainly related to air masses from southern and southwestern China), and this may reflect the spatial variations of correlation slopes in China.A seasonally dependent variation in GEM / CO correlation slopes in the upper troposphere of China was also observed (Slemr et al., 2009).It is speculated that the temporal and spatial variations of GEM / CO and GEM / CO 2 correlation slopes might be smaller than that of GEM / CH 4 , mainly due to the fact that GEM, CO, and CO 2 have many common emission sources.Nevertheless, since the correlation slopes were not obtained uniformly within different seasons and regions of the studied regions, they may be important causes of the uncertainties and limitations of the correlation slope method.Therefore, more field observations are still needed in future to better constrain Hg emissions in Asia.

Conclusions
The correlation slopes of GEM / CO, GEM / CO 2 , and GEM / CH 4 were calculated and applied for estimating the GEM emission from four important source regions in Asia using ground-based measurements at three remote sites in northwestern and southwestern China and backwards trajectory analysis.The values of GEM / CO, GEM / CO 2 , and GEM / CH 4 correlation slopes varied by source region.The GEM / CO correlation slopes were comparable among mainland China, South Asia, and the Indochinese Peninsula, with the geometric means in the range of 7.3-7.8pg m −3 ppb −1 , but they are about 2-fold lower than that in Central Asia (mean = 13.4 ± 9.5 pg m −3 ppb −1 ).This is consistent with GEM / CO 2 correlation slopes for Central Asia (mean = 315 pg m −3 ppm −1 ), South Asia (mean = 270 pg m −3 ppm −1 ), and mainland China (mean = 248 pg m −3 ppm −1 ).However, we observed an opposite spatial trend for GEM / CH 4 correlation slopes that showed the highest geometric mean of 33.3 ± 30.4 pg m −3 ppb −1 in mainland China, followed by South Asia (mean = 27.4 ± 31.0 pg m −3 ppb −1 ), the Indochinese Peninsula (mean = 23.5 ± 15.3 pg m −3 ppb −1 ), and Central Asia (mean = 20.5 ± 10.0 pg m −3 ppb −1 ).Elevated GEM / CO and GEM / CO 2 correlation slopes in Central Asia were found to be consistent with anthropogenic emission ratios of GEM relative to CO and CO 2 , indicating anthropogenic sources played an important role in the observed correlation slopes.The highest GEM / CH 4 correlation slopes in mainland China were likely due to transport from northwestern China, where strong GEM emissions and weak CH 4 emissions occur.
The observed GEM / CO, GEM / CO 2 , and GEM / CH 4 correlation slopes in Asia regions were consistently higher than those reported for Europe, North America, and South Africa.This highlights GEM emissions from non-ferrous smelting, mercury mining, natural sources, and historically deposited mercury (re-emission) in Asia.Using the correlation slopes of GEM / CO, GEM / CO 2 and recent inventories of CO and CO 2 in Asia countries, GEM emissions in mainland China, South Asia, the Indochinese Peninsula, and Central Asia were estimated to be in the ranges of 1071-1181, 340-470, 125, and 54-90 t, respectively.These estimates were lower than those predicted by the GEM / CH 4 correlation slopes because of the large uncertainties of CH 4 emissions in Asia as well as insufficient observations of GEM / CH 4 correlation slopes.These factors may lead to overestimation of GEM emissions.On the other hand, the estimates of GEM emissions in this study were much higher than those from recent anthropogenic GEM emission inventories.This discrepancy could be the result of underestimation of anthropogenic GEM emissions in Asia and GEM emissions from natural sources (including primary natural sources and re-emission of historically deposited mercury), and the uncertainties related to CO, CO 2 , and CH 4 emissions and the limitations of observed correlation slopes.Our preliminary assessment showed an annual GEM emission of about 528 t from natural sources in mainland China, and 113-240 t for South Asia, the Indochinese Peninsula, and Central Asia.Although large uncertainties exist, these estimates seem to explain the discrepancies between the calculated GEM emissions based on the observed correlation slopes and anthropogenic emissions of GEM.

Figure 1 .
Figure 1.Locations of the three ground-based remote sites in northwestern and southwestern China as well as the annual anthropogenic GEM emission for studied Asian countries (Wu et al., 2006; AMAP/UNEP, 2013).

Figure 2 .
Figure 2. A typical example of the consistent variations in GEM, CO, CO 2 , and CH 4 for the period of 30 January to 2 March 2009 at Mt. Waliguan station.

Figure 3 .
Figure 3. Histograms of the correlations slopes of GEM / CO, GEM / CO 2 , and GEM / CH 4 for the four Asian regions during the whole study period.All the correlation slopes meet the selection criteria (see text).

Figure 4 .
Figure 4. Correlation slopes of GEM / CO at the three monitoring sites and associated origins of airflows.

Figure 5 .
Figure 5. Correlation slopes of GEM / CO 2 at WLG and associated origins of airflows.

Figure 6 .
Figure 6.Correlation slopes of GEM / CH 4 at WLG and XGL and associated origins of airflows.

Fu et al.: Correlation slopes and estimated mercury emissions in ChinaTable 1 .
Statistical summary of GEM / CO, GEM / CO 2 , GEM / CH 4 , CH 4 / CO, CH 4 / CO 2 , and CO / CO 2 correlation slopes observed for mainland China, South Asia, the Indochinese Peninsula, and Central Asia during the study period.

Table 2 .
Anthropogenic emissions GEM, CO, CO 2 , and CH 4 in mainland China, South Asia, the Indochinese Peninsula, and Central Asia, as well as estimated anthropogenic emission ratios.

W. Fu et al.: Correlation slopes and estimated mercury emissions in China spectively
%, re-X.