Estimating mercury emission outflow from East Asia using CMAQ-Hg

C.-J. Lin, L. Pan, D. G. Streets, S. K. Shetty, C. Jang, X. Feng, H.-W. Chu, and T. C. Ho Department of Civil Engineering, Lamar University, Beaumont, TX 77710, USA School of Environmental Science & Engineering, South China University of Technology, Guangzhou, 510006, Guangdong, China Department of Mechanical Engineering, Lamar University, Beaumont, TX 77710, USA Decision and Information Sciences Division, Argonne National Laboratory, Argonne, IL 60439, USA Department of Chemical Engineering, Lamar University, Beaumont, TX 77710, USA Office of Air Quality Planning and Standards, USEPA, Research Triangle Park, NC 27711, USA State Key Laboratory of Environmental Geochemistry, Institute of Geochemistry, Chinese Academy of Sciences, Guiyang 550002, China

50% of all anthropogenic emissions to the atmosphere (Pacyna et al., 2006). Recently, there were a number of studies that reported experimental and modeling evidences of the long-range transport of mercury from East Asia to North America. For example, observational analysis using total mercury to carbon monoxide concentration ratio showed that the emission plumes from East Asia can be identified in the west coast 10 of North America (Jaffe et al., 2005;Weiss-Penzias et al., 2006. Global model simulations estimated that the long-range transport contributes to 5∼36% of total (e.g., dry and wet) deposition in North America depending on the locations, with an average of 16% (Jaffe and Strode, 2008;Seigneur et al., 2004). On the other hand, the impact of regional emission uncertainties (Wu et al., 2006;Shetty et al., 2008;Streets et al., 15 2005) on the source-receptor attribution estimates has not been addressed; and a detailed, quantitative assessment on the fate of atmospheric mercury in the region has not been made.
To answer the questions on how mercury emissions from East Asia may affect mercury concentration and deposition in other regions, a better understanding of the trans- 20 port, transformation and deposition in the region is needed. However, few efforts have been made to address this issue, although preliminary measurement and modeling analyses have suggested that mercury emissions transporting out of the region may be significant (Pan et al., 2006(Pan et al., , 2008Friedli et al., 2004;Weiss-Penzias et al., 2007;Kim et al., 2009). One approach to study the regional emission outflow of air pollu-In this study, the mercury model of the USEPA Community Multi-scale Air Quality modeling system (CMAQ-Hg) (Bullock and Brehme, 2002;Byun and Schere, 2006) was applied to simulate the emissions, transport, and deposition of atmospheric mercury in a model domain covering the East Asian region. The model results were incorporated in constructing the mass budget of mercury for estimating the seasonal and 5 annual mercury outflow caused by the emissions in the region. The annual outflow was estimated under three emission inventory (EI) scenarios to understand the impact of emission uncertainties. The seasonal trend of mercury chemical transport was investigated, and its implications were discussed. This work, to our knowledge, is the first modelling assessment effort on the regional chemical transport of atmospheric mer-10 cury in the region, and a part of the modeling efforts of the USEPA's Intercontinental transport and Climatic effects of Air Pollutants (ICAP) Program to understand the effect of emissions outside of the US to regional air quality via long-range transport. Interactive Discussion inferred, scaled-up emission from inverse modeling results was also considered (Pan et al., 2007) (denoted as "Inferred" case). To understand the relative importance of anthropogenic to natural source contribution, a case considering only natural and reemission was also performed (denoted as "Natural Only" case). The spatial distribution of the gridded Base-Case EI is shown in Fig. 1. The higher emission in the month of July is due to the higher surface temperature and solar radiation that drive a greater emission from natural processes. The EI for other criteria pollutants was based on the ICAP emission, which was originally based on NASA's study on the Transport and Chemical Evolution over the Pacific (Carmichael et al., 2003) with continuous updates for the 2005 modeling year.

Boundary and initial conditions
Boundary and initial conditions (BC/ICs) were re-projected from the outputs of a global 3-D chemical transport model, GEOS-Chem CTM (Selin et al., 2007) into the map projection of the study domain. To understand the effect of BCs on the estimated outflow caused by mercury emissions in the region, simulations using "background" 15 BC/ICs were also performed (details see Sect. 2.3). The background BC/ICs assumed a GEM concentration of 1.2 ng m −3 and zero concentration of RGM and PHg.

Chemical transport models
The CMAQ-Hg version 4.6 was used for all simulations. The model components, science updates, and model uncertainty issues have been discussed in details earlier 20 (Bullock and Brehme, 2002;Lin et al., 2006bLin et al., , 2007Pongprueksa et al., 2008). The Carbon Bond mechanism (CB05) was used as the gas-phase chemical mechanism to generate the concentrations of photochemical oxidants. The Rosenbrock solver (ROS3 in CMAQ CTM) was used as the chemical solver because of its flexibility (not mechanism specific). A global mass-conserving scheme (YAMO) was used for ver-Introduction Conclusions References Tables  Figures   Back  Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion was used for the vertical diffusion (documentation for these schemes is available at http://www.cmascenter.org).

Calculation of regional mercury budget
Mercury emission outflow from the domain is estimated by performing chemical transport budget calculations using the CMAQ-Hg simulation results for each modeling 5 month. A schematic for the calculation of mercury transport budget is shown in Fig. 2. The change of mercury mass within the domain boundary over a simulation period is influenced by the mercury mass entering and leaving the domain boundary, the emissions from anthropogenic and natural sources, and the deposition controlled by the chemical and physical processes in the atmosphere (e.g., chemistry, scavenging, 10 mixing, etc.). The net change within the domain boundary can be estimated as the difference between mercury masses at the beginning and at the end of the modeling period: where CM is the change of mercury mass, FM is the mercury mass in the domain at 15 the end of the modeling period (final mass), IM is the mercury mass in the domain at the beginning of the modeling period (initial mass), I is the incoming mercury mass over the modeling period, O is the outgoing mercury mass over the modeling period, E is the emission input within the domain boundary over the modeling period, and D is the removal of mercury mass by deposition in the domain over the modeling period. 20 All the terms are in Mg. Since the I and O terms represent the atmospheric transport into and out of the domain, and the transport budget (TB) can be defined as: The transport budgets for GEM, RGM and PHg were calculated for each of the modeling months. A positive value of transport budget indicates a net removal of mercury Introduction

Tables Figures
Back Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion mass in the domain (what's coming in is greater than what's going out); while a negative value indicates a net export of mercury from the domain. The mercury outflow caused (or enhanced) by the mercury emissions within the domain (OF ) can be considered as the difference in transport budget between when there is emission input and when there is no emission input, i.e., where TB i is the transport budget corresponding to one of the emission scenarios shown in Table 1, TB o is the transport budget under zero emission input. Equation (3) estimates the actual mercury emission outflow from the domain, which is independent of the BC/ICs used in the simulations. To verify that the estimated mercury outflow is 10 not influenced by BC/ICs, the outflow was estimated using two different sets of BC/ICs as described in Sect. 2.1.4. The estimated outflows (OF ) were found to be identical for both BC/IC cases.

Data analysis
The CMAQ-Hg model outputs are in network common data format (netCDF). A suite 15 of netCDF file operators developed by Zender and Mangalarn (2007) were used for the analysis of the CMAQ-Hg outputs. The Package for Analysis and Visualization of Environmental data (PAVE) version 2.3 (available at http://www.cmascenter.org) was used for data visualization.
3 Results and discussion 20

Simulated mercury concentration in East Asia and model verification
The spatial distribution of atmospheric mercury concentrations in the East Asian region was analyzed using the CMAQ-Hg model outputs of the base-case simulations (Table 1 and (Streets et al., 2005). In contrast to area and natural sources that release mercury in the surface layer only, 10 emissions from point sources have higher temperatures and are released at higher altitudes. Therefore, they have a greater potential to enter the free troposphere for long-range transport. The other feature is that there is a strong concentration gradient from the East Asian continent to the Pacific Ocean, suggesting that circumpolar westerlies transport the mercury emissions from the source region into the Pacific. 15 The simulated concentrations of atmospheric mercury in the study domain range from 1.1 to 9.3 ng m −3 , with 85-99% constituted by GEM, depending on the locations.
These results are consistent with the global model predictions (Seigneur et al., 2004;Selin et al., 2007;Strode et al., 2008). The fraction of RGM and PHg is typically greater at locations near large point sources due to the anthropogenic emission speciation 20 (nearly 50% of mercury is emitted as RGM and PHg), and decreases rapidly away from emission sources because of their relatively shorter atmospheric lifetime. The simulated surface concentrations were compared to the observed concentrations in the East Asian region (Liu et al., 2002;Wang et al., 2006Wang et al., , 2007Fang et al., 2001Fang et al., , 2004Feng et al., 2004a;Yang et al., 2009;Xiu et al., 2009;Fu et al., 2008a,b;Wan et al., ACPD 9, 21285-21315, 2009 Estimating mercury emission outflow from East Asia Interactive Discussion few years (Choi et al., 2009;Kim et al., 2009;Wan et al., 2009a,b) in the region, the model verification in terms of the magnitude of speciated mercury concentrations yields valuable insights on the model performance. The comparison is summarized in Table 2.
As seen in Table 2, the model predicted concentrations are generally consistent with 5 observations. The simulated GEM concentrations at remote sites agree excellently with field observations. However, the model under-predicted GEM concentrations in most urban areas in China by missing the peak observed concentrations. This is because a regional model such as CMAQ-Hg is incapable to simulate those high concentrations measured at the ground stations due to the model assumption of instantaneous dilution of emitted plumes in the relatively coarse grids (36-km resolution). The uncertainties of mercury emissions, which have been thought to be under-estimated or miss certain emission sources (Pan et al., 2006;Weiss-Penzias et al., 2007;Friedli et al., 2004;Feng et al., 2004aFeng et al., , 2005, may also be important reasons in the under-prediction of GEM. Model predictions of RGM and PHg agree reasonably well with the observations 15 made in China, but over-predict those observed in South Korea and at Cape Hedo, Japan (Table 2). Process analysis of model results revealed that the over-prediction of RGM and PHg in South Korea may be due to the uncertainty in the emission speciation. The somewhat over-predicted RGM and PHg at Cape Hedo was caused by the oxidation of GEM downwind of the source regions in China, and possibly the under-20 predicted dry deposition of RGM and PHg over water surface. The simulated NOx and VOCs concentration leaving the continent exhibits sufficiently strong photochemical activity to cause the oxidation, and the lower V dep of PHg compared to that of RGM leads to the greater simulation PHg concentrations at the location.

25
The spatial distribution of dry and wet deposition of atmospheric mercury of the basecase simulations with GEOS-Chem BC/ICs are shown in Figs. 4 and 5 in terms of monthly cumulated deposition fluxes (normalized to µg m −2 yr −1 ). The spatial distribu- Interactive Discussion tion of dry deposition resembles the footprint of the source locations of anthropogenic emissions (Fig. 4). The greatest deposition occurs at the immediate proximity of the emission sources, mainly in the major urban areas of China, Korea, and Japan. The simulated dry deposition is typically in the range from 10 to 150 µg m −2 yr −1 , with values exceeding 300 µg m −2 yr −1 near large point sources. The month of April appeared to 5 have a slightly larger deposition compared to other months, a result caused by a combination of relatively low planetary boundary layer height and chemical oxidation of GEM (more discussion on the domain-wide deposition in Sect. 3.3).
The spatial distribution of wet deposition is highly correlated with the locations where precipitation occurs and has a very high seasonal variability (Fig. 5). The wet depo-10 sition of mercury typically ranges from 5 to 100 µg m −2 yr −1 in the study domain. The total wet deposition is comparable but somewhat smaller than the total dry deposition (Table 3). This is consistent with the model assessments performed for North America that 30∼50% of mercury deposition is through the wet processes (Lin et al., 2007). Among the four modeling months, the month of July has the greatest wet deposition 15 because of the relatively greater precipitation and more significant chemical oxidation of GEM. The high wet deposition at the south border in July and October is caused by the high RGM concentration from the BCs in the south boundary in both months.
The scarcity of the observations for dry and wet mercury depositions limits the evaluation of model performance for the simulated mercury deposition. In China, wet

Regional mercury budget and outflow caused by the East Asian emissions
The regional mercury mass budgets of emission (E ), deposition (D), and the total mercury mass at the beginning and the end of each modeling month (IM and FM) as defined in Eq. (1) were calculated for GEM, RGM and PHg from the CMAQ-Hg model data. These values, as well as their transport budgets estimated according to Eq. (2), 5 are summarized in Table 3. From Table 3, it is clear that dry deposition is the primary removal mechanism for RGM and that wet deposition is the main removal mechanism for PHg, although the dry and wet removals are of the same order of magnitude on the four-month combined basis. GEM is primarily removed through dry deposition due to its low solubility in the aqueous phase and vegetation uptakes (Lin and Pehkonen, 1999; For the four modeling months, the transport budget (Eq. 2) of GEM is consistently negative (−128∼−62 Mg mo −1 ), while the transport budgets of RGM (7∼15 Mg mo −1 ) and PHg (13∼21 Mg mo −1 ) are consistently positive (Table 3). These indicate that there is a net mass of GEM transported out of the East Asian region and a net removal of 20 RGM and PHg in the region, consistent with earlier findings by a global model (Strode et al., 2008). The GEM mass leaving the study domain shows a strong seasonal variability and is the greatest in the month of July (128 Mg mo −1 ), mainly caused by the increased GEM emission from the natural processes. This is more than twice as much as the mass transporting out of the domain in the month of January (62 Mg mo −1 ).

25
The removal of RGM and PHg shows a weaker seasonal variability (20∼34 Mg mo −1 combined) compared to that of GEM export. The variability is mainly due to the wet deposition. The relatively smaller dry deposition variability is caused by the offset of 21296 ACPD 9, 2009 Estimating mercury emission outflow from East Asia Interactive Discussion boundary layer mixing with chemical oxidation of GEM. For example, although January has weaker GEM oxidation that leads to lower dry deposition, the shallower mixing layer height in the month causes greater RGM/PHg deposition because of less emission dilution. Summing up the transport budgets of GEM, RGM and PHg, there is an overall "export" of mercury from the East Asian region. The greatest transport quantity occurs in the month of July (99 Mg mo −1 ), more than twice as much in the month of January (42 Mg mo −1 ). On an annual basis, it is estimated that 835 Mg of mercury is transported out of the study domain, primarily in the form of GEM.
In the absence of mercury emission input, the mercury mass entering the model domain from the boundaries is readily removed due to chemical oxidation of GEM 10 followed by dry and wet deposition. The model-estimated transport budget with zero mercury emission in January, April, July and October of the modeling year is 34, 48, 50, 45 Mg, respectively (all net removal). If mercury emission is considered in the model simulations, the transport budget changed from net removal to net export. This suggests that the mercury emissions within the domain not only offset the removal 15 of incoming mercury mass from the domain boundaries, but also results in additional mercury mass leaving the domain.
The mercury outflow caused by the mercury emissions in the study domain under the three emission scenarios (Table 1) was estimated using Eq. (3). The results are shown in Fig. 6, which also exhibits a strong seasonal variability. Since the anthropogenic 20 emission inventory in the "Base Case" has been thought to be underestimated, the estimated outflow in this case can be considered a lower limit of the emission outflow. The "Inferred" case used the scaled-up emission inventory (Pan et al., 2007), therefore the estimated outflow can be considered an upper limit. Assuming that the annual outflow is three times the four-month sum, the estimated East Asian mercury outflow 25 caused by emissions is in the range of 1369∼1671 Mg yr −1 in the modeling year 2005. With the emission scenario where no anthropogenic emission was considered ("Natural Only" case), the mercury outflow caused by natural emission amounts to 805 Mg yr −1 .
The difference in the estimated annual emission outflow between the "Base/Inferred" Introduction

Implications on intercontinental and trans-boundary transport
Trans-Pacific and trans-boundary mercury transport events have been reported in a number of recent studies (Jaffe et al., 2005;Jaffe and Strode, 2008;Strode et al., 10 2008;Kim et al., 2005Kim et al., , 2009). Our earlier modeling assessment of the trans-Pacific transport of mercury showed that the direct transport of the Asian mercury plumes to North America is insignificant because the plumes have been much diluted during the process of long-range transport (Lin et al., 2006a). However, the mercury input into the atmosphere from the mercury emission outflow of the study region is significant. Interestingly, based on the monthly averaging concentration and deposition results from the CMAQ-Hg simulations performed in this study, we did not observe consistent transport events from China to Korea. This is evidenced by two 5 observations: (1) the simulated surface concentration fields do not exhibit a gradient from the industrial areas (e.g., Liaoning province) in China to the Korean region in any of the modeling months for the Base (Fig. 3) and the Inferred cases, and (2)  observations of mercury emission outflow during ACE-Asia campaign also showed that high mercury concentrations were observed at higher altitudes (>3 km above sea level) instead of near ground (Pan et al., 2006). These results imply that, even though 15 the trans-boundary mercury transport events may occur, the transport events would be episodic. The impact of such episodic transport events on dry and wet mercury depositions at the receptor sites, which is the primary concern of mercury pollution, requires further modeling assessment.

20
In this study, we investigated the regional chemical transport of atmospheric mercury in the East Asian region and the mercury emission outflow from the region using a comprehensive modeling approach coupled with regional mass balance analysis. This study marks the first regional modeling assessment in the region that accounts for nearly 50% of the global anthropogenic mercury emissions.

25
The simulated base-case mean monthly concentrations resemble the spatial distribution of GEM emissions from natural sources, with concentration hotspots caused by ACPD Introduction

Conclusions
References Tables  Figures   Back  Close Full Screen / Esc

Printer-friendly Version
Interactive Discussion large point sources. The concentration gradient from the Asian continent to the Pacific suggests significant mercury outflows. In the region, the dry deposition is mainly contributed by RGM while the wet deposition is contributed by both RGM and PHg. The annual total mercury deposition in the East Asian region for the modeling year (2005)  Feng, X. B., Wang, S. F., Qiu, G. A., Hou, Y. M., and Tang, S. L.: Total gaseous mercury emissions from soil in Guiyang, Guizhou, China, J. Geophys. Res. Atmos., 110, D14306, doi:10.1029/2004JD005643, 2005 in the atmosphere around japan, korea, and china as observed during the 2001 ace-asia 5 field campaign: Measurements, distributions, sources, and implications, J. Geophys. Res. Atmos., 109, D19S25, doi:10.1029/2003jd004244, 2004 (Wang et al., 2007); 2: