Development and Application of a Multipollutant Model for Atmospheric Mercury Deposition

By Vijayaraghavan, Krish Seigneur, Christian; Karamchandani, Prakash; Chen, Shu-Yun

ABSTRACT A multipollutant model, the Community Multiscale Air Quality model paired with the Model of Aerosol Dynamics, Reaction, Ionization, and Dissolution (CMAQ-MADRID), is extended to include a comprehensive treatment of mercury processes and is applied to the simulation of the atmospheric deposition of sulfate and mercury over the United States during 1996. Model performance is evaluated first by comparison with annual sulfate wet deposition data from the National Atmospheric Deposition Program’s National Trends Network; the coefficient of determination r^sup 2^ is 0.77, and the model normalized error and bias are 53% and -8%, respectively. When actual precipitation data are used to scale the deposition fluxes, r^sup 2^ improves to 0.91 and the error and bias change to 42% and -41%, respectively. The scaled results underscore a tendency of the model to underestimate sulfate wet deposition. Model performance for mercury wet deposition is then evaluated by comparison with data from the Mercury Deposition Network. For annual mercury wet deposition, r^sup 2^ is 0.28 and the normalized error and bias are 81% and 73%, respectively, when the modeled precipitation data are used. Model performance improves when actual precipitation data are used to scale deposition fluxes: r^sup 2^ increases to 0.41 and the error and bias decrease to 40% and 29%, respectively. The model reproduces the spatial pattern of sulfate wet deposition adequately with an increasing gradient from the upper Midwest to the Northeast, that is, from upwind to downwind of large sulfur dioxide sources in the Ohio River Valley. However, the model tends to overestimate mercury wet deposition in the Northeast downwind of these sources that also emit significant amounts of mercury. This “Pennsylvania anomaly” may be due to a partial misrepresentation of the mercury reduction-oxidation cycle, uncertainties in the dry deposition of divalent gaseous mercury Hg^sup II^, incorrect speciation of mercury emissions, and/or uncharacterized emissions in the upper Midwest.

1. Introduction

Deposition of atmospheric mercury (Hg) to water bodies and their watersheds may lead to the bioaccumulation of methyl-Hg in the aquatic food chain and the potential exposure of humans and animals consuming fish. Because methyl-Hg may be harmful at high levels, it is important to understand the physicochemical processes that govern the overall cycle of Hg in the environment. Therefore, a large amount of work has been carried out to construct, evaluate, and apply mathematical models of the atmospheric fate and transport of Hg. In particular, several models have been applied to Hg deposition in North America (Pai et al. 1997; Xu et al. 2000; Bullock and Brehme 2002; Lin and Tao 2003; Cohen et al. 2004; Seigneur et al. 2004). Model simulations to date have focused mostly on Hg with other pollutants being input to the model as reactants with predetermined concentrations. Two exceptions include the versions of the San Joaquin Valley Air Quality Study and Atmospheric Utility Signatures Predictions and Experiments Study Regional Modeling Adaptation Project (SARMAP) Air Quality Model (SAQM; Xu et al. 2000) and of the Community Multiscale Air Quality (CMAQ) model (Bullock and Brehme 2002) that include Hg processes. SAQM and CMAQ treat, in addition to Hg, all major gas-phase air pollutants, including nitrogen oxides (NO^sub x^), volatile organic compounds (VOC), carbon monoxide (CO), and sulfur dioxide (SO2). CMAQ also treats particulate matter (PM) and atmospheric deposition; it is, therefore, considered to be a “one atmosphere” model because it can simulate the major forms of urban and regional air pollution. Both SAQM and CMAQ have been applied to short periods (from one week to a few weeks) over the eastern United States (Xu et al. 2000; Bullock and Brehme 2002; Lin and Tao 2003). CMAQ was recently applied for a full year over the contiguous United States (EPA 2005).

The use of one-atmosphere models is of particular interest because emission control systems typically affect the emissions of several pollutants. For example, in the case of coal-fired power- plant emissions, some particulate-bound Hg (Hg^sub p^) is removed by PM control equipment (Pavlish et al. 2003). Flue gas desulfurization systems (FGD) that are used to remove SO2 will also remove divalent gaseous Hg (Hg^sup II^) very effectively (Pavlish et al. 2003; Niksa and Fujiwara 2005a) because Hg^sup II^ is mostly present as HgCl2 (Cl is chlorine) in power-plant flue gases (Senior et al. 2000), and this Hg species is very water soluble (Lindqvist and Rodhe 1985); elemental mercury (Hg^sup 0^) is not removed because it is relatively insoluble (Sanemasa 1975). Selective catalytic reduction systems (SCR) that are used to remove NO^sub x^ tend to oxidize Hg^sup 0^ to Hg^sup 11^ (Richardson et al. 2002; Eswaran and Stenger 2005; Niksa and Fujiwara 2005b; Senior 2006); as a consequence, an SCR located upstream of an FGD will convert some Hg^sup 0^ to Hg^sup 11^ that can then be captured by the FGD. In summary, emission control systems tend to be effective for several pollutants, and it is desirable to use one-atmosphere models that can treat all relevant air pollutants simultaneously to analyze the full effects of emission control strategies on air quality.

We present here the extension of a one-atmosphere model, CMAQ paired with the Model of Aerosol Dynamics, Reaction, lonization, and Dissolution (MADRID), to assess mercury processes. The model is evaluated for two major pollutants of concern for atmospheric deposition to watersheds: sulfate (SO^sup 2-^^sub 4^) and mercury. Coal-fired power plants emit, in the absence of specific emission control equipment, both SO2 (the precursor of sulfate) and Hg; it is of interest, therefore, to compare the deposition patterns of sulfate and mercury upwind and downwind of these sources. The modeling of sulfate deposition dates back to the 1980s when SO2 emission control strategies were first developed to reduce acidic deposition in North America (Irving 1991). As a result, there exist both an extensive database for atmospheric sulfate wet deposition data in the National Atmospheric Deposition Program (NADP) National Trends Network (NTN; NADP-NTN 2006) and a significant historical set of model performance evaluations for sulfate deposition (e.g., Karamchandani et al. 2002). Thus, evaluating CMAQ-MADRID for sulfate allows us to perform a more comprehensive performance evaluation than is possible for mercury because the NADP’s Mercury Deposition Network (MDN; NADP-MDN 2006) is less dense spatially than the sulfate NADP network and has been in operation for a shorter time. Pai et al. (1997) used a similar twostep approach in their evaluation of the Trace Element Analysis Model (TEAM). Thus, we first evaluate CMAQ-MADRID for sulfate wet deposition. Next, we evaluate the model for mercury wet deposition. The ability of the model to reproduce the spatial patterns of atmospheric deposition observed in the NADP is then analyzed, and major differences between sulfate and mercury deposition patterns are discussed.

2. Description of the one-atmosphere model

CMAQ is a three-dimensional (3D) air-quality model that simulates ozone and other photochemical oxidants, PM, and the deposition of pollutants such as acidic and nitrogenous compounds (Byun and Ching 1999). CMAQ-MADRID refers to a version in which MADRID (Zhang et al. 2004) is used instead of the standard CMAQ PM module to simulate the formation of PM. The version of CMAQ-MADRID used in this study is based on CMAQ, version 4.4.

The atmospheric lifetime of Hg is on the order of 1 yr (Seigneur et al. 2004). Note, however, that Hg can cycle several times between the Hg^sup 0^ and Hg^sup 11^ species before being removed from the atmosphere and that the atmospheric chemical lifetimes of the individual species Hg^sup 0^ and Hg^sup 11^ are significantly less than 1 yr. The Hg chemical kinetic mechanism and deposition processes that are treated in TEAM (Seigneur et al. 2004) were incorporated into CMAQ-MADRID. There are uncertainties associated with our current understanding of the atmospheric chemistry of Hg, including the kinetics of the oxidation of Hg^sup 0^ by ozone (O3) and hydroxyl radical (OH) (Calvert and Lindberg 2005), the reduction of Hg^sup 11^ (Gardfeldt and Jonsson 2003), and the speciation of Hg in power-plant plumes (Lohman et al. 2006). Therefore, the mechanism used here is likely to evolve as new laboratory data on the chemical kinetics and thermodynamics of Hg species become available. Nevertheless, this current chemical kinetic mechanism provides a representation of the reduction and oxidation pathways of Hg species that is consistent with the global Hg cycle (Seigneur et al. 2006). The options of CMAQ-MADRID used with the Hg formulation were selected to provide a balance between scientific accuracy and computational efficiency. Computational efficiency is important because Hg simulations are typically performed for North America for an entire year. The major modules used in this extended model, CMAQ- MADRID-Hg, are briefly described below. The solution of the advection processes (horizontal and vertical advection) is obtained with the piecewise parabolic method (Colella and Woodward 1984). Horizontal diffusion is simulated with a scheme that is a function of the grid size (material diffusion is decreased as the grid size increases to account for increased numerical diffusion). Vertical diffusion is simulated with eddy-diffusion algorithms based on similarity theory (Businger et al. 1971; Chang et al. 1987; Byun and Dennis 1995).

Version 4 of the Carbon Bond Mechanism (CBMIV) is used to simulate the gas-phase chemistry of NO^, VOC, SO2, and CO (Gery et al. 1989). The numerical solution is obtained with the Euler backward iterative solver (Hertel et al. 1993). The Regional Acid Deposition Model (RADM) chemical mechanism is used to simulate aqueous-phase chemistry in clouds (Walcek and Taylor 1986).

The formation of PM is simulated with MADRID (Zhang et al. 2004). Inorganic PM includes sulfate, nitrate, ammonium, sodium, chloride, and water. The model also tracks elemental carbon and primary organic aerosol. secondary organic aerosol formation is simulated from the oxidation of two anthropogenic precursors and six biogenic precursors (Pun et al. 2004). The particle size distribution is represented by a sectional distribution. Two size sections (fine and coarse) are used for most applications.

Dry deposition is simulated using a resistance transfer approach for gases and particles (Pleim et al. 1997). For particles, the formulation of Venkatram and Pleim (1999) is used to account for gravitational settling. For Hg^sup 0^, background emissions and dry deposition are assumed to balance each other over North America. This common assumption (e.g., Bullock and Brehme 2002) is justified by the fact that the atmospheric lifetime of Hg^sup 0^ (several months) greatly exceeds its residence time (a few days) within the modeling domain; therefore, its concentration within the domain is not very sensitive to its removal rate. Dry deposition characteristics of Hg^sup II^ are assumed to be similar to those of nitric acid because both gases have similar solubilities in water. The dry deposition of Hg^sub p^ is treated similarly to that of fine particles.

Clouds are obtained from the output of a prognostic meteorological simulation conducted with the fifthgeneration Pennsylvania State University-National Center for Atmospheric Research Mesoscale Model (MM5; Grell et al. 1994). Clouds that are subgrid scale are predicted with the technical approach used in RADM (Chang et al. 1987; Dennis et al. 1993). Wet deposition includes rainout and washout. Rainout is calculated using the chemical concentrations in the cloud droplets calculated from the aqueous chemistry and the precipitation rate. A cloud nuclei activation algorithm is used to estimate the scavenging of particles in clouds. Washout of gases (i.e., scavenging of chemical species below the cloud base by precipitation) is calculated using an estimate of the dissolution of gases in raindrops (based on their effective Henry’s law coefficients). Scavenging coefficients are used for the washout of particles.

There are, therefore, several differences between CMAQ-MADRID (this work) and CMAQ (Bullock and Brehme 2002), which can be summarized as follows. Both models use the same major reactions (with the same chemical kinetics) for Hg^sup 0^ oxidation and Hg^sup 11^ reduction; there are, however, some minor differences, such as the treatment of the oxidation product of the reaction of Hg^sup 0^ with O3 (gaseous HgO in CMAQ-MADRID and a mixture of Hg^sub p^ and Hg^sup II^ in CMAQ) and the treatment of Hg^sub p^ in droplets (insoluble in CMAQ-MADRID and soluble in CMAQ). A comparison of the two chemical kinetic mechanisms showed some differences but similar qualitative behavior when predicting the Hg oxidation-reduction cycle (Ryaboshapko et al. 2002) (the gas-phase oxidation of Hg^sup 0^ by OH was added to both models after this comparative study). The treatment of aerosol processes differs between the two models; in particular, CMAQ-MADRID uses a sectional representation of the particle size distribution, whereas CMAQ uses a modal representation. This difference affects the dry deposition velocity of Hgp, which is associated with fine particles in both models, and, to some minor extent, Hg^sub p^ wet scavenging in clouds because the cloud droplet activation algorithms differ in the two models. Other major differences in the treatment of aerosol processes (e.g., formation of secondary organic aerosols and heterogeneous reactions on particles) do not directly affect Hg processes and should have negligible impacts on Hg transformations and deposition. Differences in the model inputs include the Cl2 concentration fields (prescribed as a 3D concentration field in CMAQ-MADRID and simulated from Cl2 emissions in CMAQ) and the boundary conditions, which are obtained from different global models [the Mercury Chemical Transport Model (CTM-Hg) for CMAQ-MADRID (see below) and the Goddard Earth Observing System 3D atmospheric composition model (GEOS-Chem) for CMAQ].

3. Application of the model to the continental United States

CMAQ-MADRID-Hg was applied to simulate sulfate and Hg deposition over the contiguous United States for 1996. This year was selected because it was the most recent year for which MM5 output results were available from the U.S. Environmental Protection Agency (EPA) at the time of this work. The domain encompasses the contiguous United States and parts of Canada and Mexico. The modeling grid is defined in a Lambert conical projection with reference latitudes at 30[degrees] and 60[degrees]N, the central meridian at 100[degrees]W, and the coordinate system origin at 40[degrees]N, 100[degrees]W. The southwest corner of the grid is 2196 km west and 1692 km south of the coordinate system origin. The grid domain contains 132 x 90 grid cells with 36-km horizontal spacing. The vertical resolution for CMAQ-MADRID-Hg includes 12 layers from the surface to the tropopause (approximately 15-km altitude), corresponding to sigma levels of 1.000, 0.995, 0.988, 0.970, 0.938, 0.893, 0.839, 0.777, 0.702, 0.582, 0.400, 0.200, and 0.000 at the layer boundaries {sigma is defined as sigma = [(p - p^sub top^)/ (P^sub s^ - P^sub top^)]. where p^sub top^ is a defined upper pressure level, p^sub s^ is the pressure at the (geometrically variable) surface, and p is pressure at the intermediate level). Sigma coordinates are terrain following. The surface layer extends up to a height of roughly 35 m above terrain.

Daily 3D gridded emissions of VOC, NO^sub x^, ammonia (NH^sub 3^), SO^sub 2^, CO, and PM for 1996 were obtained from EPA (N. Possiel 2003, personal communication). The PM size distributions provided by EPA according to three modes were converted to two size sections: the Aitken and accumulation modes were assigned to the fine PM size section and the coarse mode was assigned to the coarse PM size section. A speciated emissions inventory developed for 1998- 99 was used for Hg (Pai et al. 2000; Seigneur et al. 2004). Note that some waste incinerator emissions decreased from 1996 (the simulation year) to 1998; this fact was not taken into account in this study. All particulate-bound Hg emissions were assumed to be associated with fine PM (i.e., the first size section).

MM5 meteorological fields for 1996 were obtained from EPA and were processed with the MeteorologyChemistry Interface Processor, version 2.2 (MCIP 2.2). The results of a global mercury model simulation (Seigneur et al. 2004) were used to provide the boundary conditions for the CMAQ-MADRID-Hg application to North America. The global model uses the same chemical kinetic mechanism as CMAQ- MADRID-Hg and includes algorithms for wet and dry deposition that are simpler than but consistent with those used in CMAQ-MADRID-Hg. Thus, the boundary conditions are obtained from a model that is consistent with CMAQMADRID-Hg in its treatment of Hg processes. These boundary conditions consist of concentrations of Hg^sup 0^, Hg^sup II^, and Hg^sub p^ as a function of location, height, and season. The global grid cells used for setting boundary conditions range from 20[degrees] to 68[degrees]N latitude and from 45[degrees] to 145[degrees]W longitude; because the global model uses a grid spacing of 8[degrees] latitude and 10[degrees] longitude, several grid cells provide spatially varying concentrations for each boundary. The global model includes nine layers, up to an altitude of about 31 km. The first seven of these layers extend to an altitude of approximately 14 km and were mapped into the 12 layers of CMAQ-MADRID-Hg. The boundary conditions vary according to season. The values simulated by the global model for January, April, July, and October were used to represent winter, spring, summer, and autumn conditions, respectively (e.g., the monthly average January concentrations from the global model provided the boundary conditions for December, January, and February in the continental simulation; the monthly average global concentrations for April were used for March, April, and May; and so on). Thus, the global CTM provides spatially distributed and temporally resolved fields of background Hg species concentrations as boundary conditions for CMAQ- MADRID-Hg. Figure 1 shows the average vertical distribution of mercury species concentrations in January (winter) and July (summer) at the four boundaries of the modeling domain. Default vertical profiles provided by EPA that do not vary in time were used for initial and boundary conditions of modeled species other than Hg. The annual simulation was performed as four separate 3-month runs, each with a 10-day model spinup period. Concentrations of molecular chlorine (Cl2) and hydrogen chloride (HCl), which are not simulated by the CMAQ chemical kinetic mechanism, were prescribed using the approach adopted by Seigneur et al. (2004). 4. Performance evaluation for sulfate

Figure 2 presents the simulated total (i.e., wet plus dry) annual deposition of sulfate over the contiguous United States. The highest sulfate deposition fluxes occur mostly in the northeastern United States, ranging up to 50 kg ha^sup -1^ yr^sup -1^. In 1996, atmospheric wet deposition of sulfate was measured at 189 NADP-NTN monitoring stations in the contiguous United States (dry deposition is not measured in the NADP network). Figure 3 presents a comparison of maps of measured and modeled sulfate wet deposition. (The top panel of Fig. 3 resembles Fig. 2 because sulfate dry deposition is small relative to sulfate wet deposition.) The major spatial patterns of sulfate wet deposition are simulated well by the model. Measured annual sulfate wet deposition is less than 15 kg ha^sup – 1^ west of the Mississippi River, except for the eastern parts of Missouri, Arkansas, and Louisiana. It is also less than 15 kg ha^sup -1^ in most of the Southeast (Florida, southern Georgia, and southern South Carolina) and in the upper Northeast (e.g., Maine). The modeled values show similar patterns, with slightly greater wet deposition values around the Four Corners area in the Southwest and in northeastern Texas. Measured sulfate wet deposition fluxes exceed 20 kg ha^sup -1^ yr^sup -1^ in the Northeast, with the highest values around the Ohio River Valley and Pennsylvania; similar fluxes are simulated by the model. Note that the spatial interpolation used to create the MDN maps may not be completely accurate in regions with sparse sampling sites; this situation could affect the intercomparison with model results in the western United States.

Figure 4 presents scatterplots of the annual measured and simulated sulfate wet deposition fluxes at all NADP-NTN sites in 1996. When the precipitation amounts simulated by MM5 are used, the coefficient of determination r^sup 2^ is calculated to be 0.77. The mean normalized error and bias are 53% and -8%, respectively. The definition of these statistical measures has been provided by Seigneur et al. (2000). If one uses the actual precipitation amounts measured at the NADP sites to scale the simulated sulfate wet deposition amounts, the following statistics are obtained: r^sup 2^ is 0.91, and the mean normalized error and bias are 42% and -41%, respectively. These scaled results underscore a tendency of the model to consistently underestimate sulfate wet deposition: this tendency is demonstrated by the values of absolute error and bias, which are similar in value but of opposite sign, and the relatively high coefficient of determination. Notwithstanding, model performance would likely improve if NADP or National Weather Service precipitation observations were used to correct MM5 precipitation fields input to the model rather than being used for scaling output deposition fluxes as a postprocessing step (Vijayaraghavan et al. 2005).

5. Performance evaluation for mercury

Figure 5 presents simulated Hg annual deposition fluxes for total (i.e., wet plus dry) deposition, wet deposition and dry deposition, respectively, of Hg^sup II^ and Hg^sub p^ (Hg^sup 0^ dry deposition is not calculated because most drydeposited Hg^sup 0^ is assumed to be reemitted; note that this approximation may not be true for water bodies for which some of the reemitted Hg^sup 0^ has been formed by the reduction of deposited Hg^sup II^ in the water column). The Hg deposition is greater on average in the eastern United States than in the western United States, although there are some exceptions. For example, Hg deposition shows higher values in the Geyser resource area in northern California when compared with other West Coast locations because of Hg emissions associated with geothermal activity. In the eastern United States, the region around the Ohio River Valley and the area downwind (e.g., Pennsylvania) tend to show the highest simulated total mercury deposition fluxes (greater than 75 [mu]g m^sup -2^ yr^sup -1^).

Annual Hg wet deposition fluxes simulated with CMAQ-MADRID-Hg were compared with measurements at the NADP-MDN stations for which data were available for 1996. Figure 6 presents a scatterplot of the measured and simulated Hg annual wet deposition fluxes at the 11 MDN sites available for 1996. The model explains only 28% of the variance in the data and it shows a strong tendency to overpredict. An analysis of the results shows that a significant fraction of the overprediction is due to an overprediction of the annual precipitation by MM5 in some locations (the largest overpredictions occur in North Carolina, South Carolina, and Texas).

Figure 6 also shows a scatterplot after the simulated wet deposition results have been scaled by the ratio of observed to simulated precipitation amounts. The correlation improves with a value of r^sup 2^ of 0.41, and the overprediction has been significantly reduced. CMAQ-MADRID-Hg shows a positive bias of 29% and an absolute error of 40%. The correlation obtained with CMAQ- MADRID-Hg in this 1996 simulation is lower than that obtained with TEAM in a 1998 simulation (r^sup 2^ = 0.50; Seigneur et al. 2004), primarily because no MDN sites were available in 1996 in the western United States where mercury deposition is typically lower than it is in the eastern United States. The inclusion of sites with lower mercury wet deposition would tend to increase the correlation between measured and simulated values (assuming satisfactory model performance at such sites).

There were not enough MDN sites in 1996 to allow a comparison of the spatial distribution of measured and modeled mercury wet deposition as was done for sulfate in Fig. 3. Therefore, spatial patterns of measured mercury wet deposition using more recent data from 2004 are discussed below.

6. Comparison of spatial patterns of sulfate and mercury wet deposition

Although mercury wet deposition data are limited for 1996 (i.e., the model simulation year), some clear differences appeared between the spatial patterns of the wet deposition of sulfate (Fig. 3) and Hg (Fig. 5). It is, therefore, of interest to examine whether such differences are confirmed with data from recent years when a more extensive MDN database is available. First, we present in Fig. 7 a comparison between measured annual wet deposition fluxes of mercury and sulfate in 1996 and 2004 at the MDN sites where data are available for both years (11 such sites existed for mercury, and there were 181 for sulfate). There is a good correlation between the 1996 and 2004 MDN wet deposition fluxes (r^sup 2^ = 0.82 and 0.86 for Hg and SO^sup 2-^^sub 4^, respectively), thus indirectly validating the extension of our analysis to 2004. Figure 8 presents a comparison of wet deposition data for sulfate and Hg in 2004. The amount of precipitation affects the wet deposition flux; therefore, the wet deposition fluxes and concentrations in precipitation are both presented (concentrations in precipitation are less sensitive to the precipitation amount, although a larger precipitation amount can lead to depletion of a soluble pollutant and consequent dilution of its concentration in precipitation).

Sulfate wet deposition fluxes have noticeably declined from 1996 to 2004 as a result of reductions in SO^sub 2^ source emissions, and the area in which sulfate wet deposition exceeds 20 kg ha^sup -1^ yr^sup -1^ is significantly smaller in 2004 than in 1996. Nevertheless, as for the 1996 results, the 2004 sulfate spatial maps for both wet deposition and concentrations in precipitation show larger values in the Northeast and lower values in the West and the Southeast. The gradient from Minnesota to the Northeast is explained by the presence of several large SO^sub 2^ sources in the general area of the Ohio River Valley.

On the other hand, the spatial map for Hg wet deposition presents a weaker gradient from Minnesota to Pennsylvania than that observed for sulfate. This “Pennsylvania anomaly” is more evident when comparing concentrations of Hg and sulfate in precipitation; the gradient in wet deposition is governed by a gradient in precipitation (not shown; see NADP-NTN 2006), and concentrations of Hg in precipitation actually show slightly lower values in the Northeast than in Minnesota or Wisconsin. It is unlikely that the lower Hg concentrations in the Northeast are due solely to dilution associated with greater precipitation because such dilution does not appear in the sulfate concentrations in precipitation (see Fig. 8). On the other hand, the Hg wet deposition fluxes show a clear north- to-south gradient, with greater deposition fluxes in Florida, Louisiana, and Mississippi than in the Northeast. This discrepancy between sulfate and Hg spatial deposition patterns may be due to differences in their chemistry, dry deposition, and/or emission sources.

The oxidation of SO^sub 2^ to sulfate occurs rapidly in clouds, and, as a result, sulfate deposition tends to occur at regional scales (on the order of hundreds of kilometers) downwind of SO^sub 2^ sources. On the other hand, the oxidation of Hg^sub 2^ to Hg^sup II^ is slower (with a half-life on the order of months) and reduction of Hg^sup II^ to Hg^sub 0^ may also take place. In particular, there is some evidence that Hg^sup II^ may be reduced to Hg^sup 0^ in some power-plant plumes (Edgerton et al. 2004; Prestbo et al. 2004; Lohman et al. 2006). Dry deposition of SO^sub 2^ and sulfate particles is most likely slower than dry deposition of Hg^sup II^ (the Hg species that dominates Hg wet deposition) (Seinfeld and Pandis 1998; Lindberg and Stratton 1998; Poissant et al. 2004) and could affect wet deposition patterns to some extent because sulfur (S) and Hg species removed by dry deposition will not be available for wet deposition. Incorrect speciation of mercury emissions from coal-fired power plants could also contribute to the anomaly; overestimation of the Hg^sup II^ fraction in emissions would result in overprediction of mercury deposition downwind of those sources. Last, although both S and Hg are emitted from coal- fired power plants, there are other sources of S and Hg that may differ in their relative spatial distributions. Previous work has shown that models of atmospheric mercury deposition do not reproduce the spatial mercury deposition patterns well in the Northeast (Seigneur et al. 2003)-in particular, in Pennsylvania. Because the one-atmosphere model used here was able to properly reproduce spatial deposition patterns for sulfate, we propose that some processes that govern the atmospheric concentrations of mercury are not yet adequately represented in models.

7. Conclusions

The development and initial application of a oneatmosphere model for the atmospheric deposition of mercury is presented. This approach is useful because emission control systems used to reduce emissions of PM, SO2, and NO^sub x^ may also reduce Hg emissions from those same sources. Thus, a one-atmosphere model offers the possibility of assessing the effect of emission control strategies on multiple pollutants.

Because Hg data are still limited-in particular, in the western United States-the model was first evaluated for sulfate wet deposition. Model performance was satisfactory, demonstrating the model’s ability to simulate well the major transport, dispersion, and deposition processes. Evaluation of model performance for Hg revealed the sensitivity of calculated wet deposition fluxes to the simulated precipitation amounts. Using precipitation observations to scale deposition fluxes significantly improved model performance.

A comparison of sulfate and mercury deposition maps for 2004 illustrates that these two pollutants show different spatial deposition patterns. In particular, Hg deposition in the Northeast is not much greater than that in the upper Midwest, and the east- west gradient is due to precipitation amounts rather than Hg concentrations in precipitation. On the other hand, sulfate wet deposition shows a clear gradient from the upper Midwest to the Northeast. These differences between the spatial patterns of the two pollutants may result from differences in their chemistry, dry deposition, and/or emission patterns.

Acknowledgments. This work was presented at the CMAS Conference in Chapel Hill, North Carolina, 2628 September 2005. The incorporation of mercury in CMAQ-MADRID and its application to sulfate and mercury deposition were funded under Contract EPP15776/ C7851 by EPRI, Palo Alto, California; the development of CMAQ- MADRID has been funded by EPRI under Contract EP-P15051/C7419. We thank the EPRI project managers, Drs. Leonard Levin and Eladio Knipping, for their continuous support.

REFERENCES

Bullock, O. R., and K. A. Brehme, 2002: Atmospheric mercury simulation using the CMAQ model: Formulation description and analysis of wet deposition results. Atmos. Environ., 36, 2135-2146.

Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F. Bradley, 1971: Flux profile relationships in the atmospheric surface layer. J. Atmos. Sci., 28, 181-189.

Byun, D. W., and R. L. Dennis, 1995: Design artifacts in Eulerian air quality models: Evaluation of the effects of layer thickness and vertical profile correction on surface ozone concentrations. Atmos. Environ., 29, 105-126.

_____, and J. K. S. Ching, 1999: Science algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) modeling system. U.S. Environmental Protection Agency Scientific Doc. EPA/600/R-99/ 030, 706 pp.

Calvert, J. G., and S. E. Lindberg, 2005: Mechanisms of mercury removal by O3 and OH in the atmosphere. Atmos. Environ., 39, 3355- 3367.

Chang, J. S., R. A. Brest, I. S. A. Isaksen, S. Madronich, P. Middleton, W. R. Stockwell, and C. J. Walcek, 1987: A threedimensional Eulerian acid deposition model: Physical concepts and formulation. J. Geophys. Res., 92, 14 681-14 700.

Cohen, M., and Coauthors, 2004: Modeling the atmospheric transport and deposition of mercury to the Great Lakes. Environ. Res., 95, 247-265.

Colella, P., and P. R. Woodward, 1984: The piecewise parabolic method (PPM) for gas-dynamical simulations. J. Comput. Phys., 54, 174-201.

Dennis, R. L., J. N. McHenry, W. R. Barchet, F. S. Binkowski, and D. W. Byun, 1993: Correcting RADM’s sulfate underprediction: Discovery and correction of model errors and testing the corrections through comparisons against field data. Atmos. Environ., 26A, 975- 997.

Edgerton, E. S., J. Jansen, and B. E. Hartsell, 2004: Speciated Hg measurements at a rural site near Atlanta, GA USA. RMZ Mater. Geoenviron., 51, 1539.

EPA, 2005: Air quality modeling: Technical support document for the final Clean Air Mercury Rule: Air quality modeling. Office of Air Quality Planning and Standards, 22 pp.

Eswaran, S., and H. G. Stenger, 2005: Understanding mercury conversion in selective catalytic reduction (SCR) catalysts. Energy Fuels, 19, 2328-2334.

Gardfeldt, K., and M. Jonsson, 2003: Is bimolecular reduction of Hg(II)-complexes possible in aqueous systems of environmental importance? J. Phys. Chem., 107, 4478-4482.

Gery, M. W., G. Z. Whitten, J. P. Killus, and M. C. Dodge, 1989: A photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94,12 925-12 956.

Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 122 pp.

Hertel, O., R. Berkowicz, J. Christensen, and O. Hov, 1993: Test of two numerical schemes for use in atmospheric transportchemistry models. Atmos. Environ., 27A, 2591-2611.

Irving, P.M., Ed., 1991: Emissions, Atmospheric Processes, and Deposition. Vol. 1, Acidic Deposition: State of Science and Technology, National Acid Precipitation Assessment Program, 1760 pp.

Karamchandani, P., C. Seigneur, and Y. Zhang, 2002: Review of air quality models for acid deposition. U.S. Environmental Protection Agency Tech. Doc. EPA-430R-02-001, 183 pp.

Lin, X., and Y. Tao, 2003: A numerical modeling study on regional mercury budget for North America. Atmos. Chem. Phys., 3, 535-548.

Lindberg, S. E., and W. J. Stratton, 1998: Atmospheric mercury speciation: Concentrations and behavior of reactive gaseous mercury in ambient air. Environ. Sd. Technol, 32, 49-57.

Lindqvist, O., and H. Rodhe, 1985: Atmospheric mercury: A review. Tellus, 37B, 136-159.

Lohman, K., C. Seigneur, E. Edgerton, and J. Jansen, 2006: Modeling mercury in power plant plumes. Environ. Sci. Technol., 40, 3848-3854.

NADP-MDN, cited 2006: MDN objectives. Mercury Deposition Network: An NAPD Network. [Available online at http:// nadp.sws.uiuc.edu/ mdn.]

NADP-NTN, cited 2006: NTN sites. [Available online at http:// nadp.sws.uiuc.edu.]

Niksa, S., and N. Fujiwara, 2005a: The impact of wet FGD scrubbing on Hg emissions from coal-fired power stations. J. Air Waste Manage. Assoc., 55, 970-977.

_____, and _____, 2005b: A predictive mechanism for mercury oxidation on selective reduction catalysts under coal-derived flue gas. J. Air Waste Manage. Assoc., 55, 1866-1875.

Pai, P., P. Karamchandani, and C. Seigneur, 1997: Simulation of the regional atmospheric transport and fate of mercury using a comprehensive Eulerian model. Atmos. Environ., 31, 27172732.

_____, D. Niemi, and B. Powers, 2000: A North American inventory of anthropogenic mercury emissions. Fuel Process. Technol., 65-66, 101-115.

Pavlish, J. H., S. E. Sondreal, M. D. Mann, E. S. Olson, K. C. Galbreath, D. L. Laudal, and S. A. Benson, 2003: Status review of mercury control options for coal-fired power plants. Fuel Process. Technol., 82, 89-165.

Pleim, J. E., A. Xiu, P. L. Finkelstein, and J. F. Clarke, 1997: Evaluation of a coupled land-surface and dry deposition model through comparison to field measurements of surface heat, moisture, and ozone fluxes. Proc. 12th Symp. on Boundary Layers and Turbulence, Vancouver, BC, Canada, Amer. Meteor. Soc., 478-479.

Poissant, L., M. Pilote, X. Xu, H. Zhang, and C. Beauvais, 2004: Atmospheric mercury speciation and deposition in the Bay St. Francois wetlands. J. Geophys. Res., 109, D11301, doi:10.1029/ 2003JD004364.

Prestbo, E. M., and Coauthors, 2004: Interconversion of emitted atmospheric mercury species in coal-fired power plant plumes. RMZ Mater. Geoenviron., 51, 1732-1733.

Pun, B. K., C. Seigneur, and E. Knipping, 2004: Optimizing secondary organic aerosol representation in paniculate matter air quality models. Extended Abstract, Air & Waste Management Association Visibility Specialty Conf.-Regional and Global Perspectives on Haze: Causes, Consequences and Controversies, Asheville, NC, Air and Waste Management Association, Paper No. 14.

Richardson, C, T. Machalek, S. Miller, C. Dene, and R. Chang, 2002: Effect of NO^sub x^ control processes on mercury speciation in utility flue gas. J. Air Waste Manage. Assoc., 52, 941-947.

Ryaboshapko, A., and Coauthors, 2002: Comparison of mercury chemistry models. Atmos. Environ., 36, 3881-3898.

Sanemasa, I., 1975: The solubility of elemental mercury vapor in water. Bull. Chem. Soc. Japan, 48, 1795-1798.

Seigneur, C., and Coauthors, 2000: Guidance for the performance evaluation of three-dimensional air quality modeling systems for particulate matter and visibility. J. Air Waste Manage. Assoc., 50, 588-599.

_____, P. Karamchandani, K. Vijayaraghavan, R.-L. Shia, and L. Levin, 2003: On the effect of spatial resolution on atmospheric mercury modeling. Sd. Total Environ., 304, 73-81.

_____, K. Vijayaraghavan, K. Lohman, P. Karamchandani, and C. Scott, 2004: Global source attribution for mercury deposition in the United States. Environ. Sci. Technol., 38, 555-569.

_____, _____, and _____, 2006: Atmospheric mercury chemistry: Sensitivity of global model simulations to chemical reactions. J. Geophys. Res., Ill, D22306, doi: 10.1029/2005JD006780.

Seinfeld, J. H., and S. N. Pandis, 1998: Atmospheric Chemistry and Physics. John Wiley and Sons, 1326 pp. Senior, C. L., 2006: Oxidation of mercury across selective catalytic reduction catalysts in coal-fired power plants. J. Air Waste Manage. Assoc., 56, 23-31.

_____, A. F. Sarofim, T. Zeng, J. J. Helble, and R. Mamani-Paco, 2000: Gas-phase transformations of mercury in coal-fired power plants. Fuel Process. Technol., 63, 197-213.

Venkatram, A., and J. Pleim, 1999: The electrical analogy does not apply to modeling dry deposition of particles. Atmos. Environ., 33, 3075-3076.

Vijayaraghavan, K., C. Seigneur, P. Karamchandani, K. Lohman, and S.-Y. Chen, 2005: Modeling of atmospheric mercury deposition in the United States. Proc. NADP 2005 Technical Committee Meeting, Jackson, WY, National Atmospheric Deposition Program, p. 62.

Walcek, C. J., and G. R. Taylor, 1986: A theoretical method for computing vertical distribution of acidity and sulfate production within cumulus clouds. J. Atmos. Sci., 43, 339-355.

Xu, X., X. Yang, D. Miller, J. J. Heible, and R. J. Carley, 2000: A regional scale modeling study of atmospheric transport and transformation of mercury, 1. Model development and evaluation. Atmos. Environ., 34, 4933-4944.

Zhang, Y., and Coauthors, 2004: Development and application of the Model of Aerosol Dynamics, Reaction, lonization and Dissolution (MADRID). J. Geophys. Res., 109, D01202, doi:10.1029/2003JD003501.

KRISH VIJAYARAGHAVAN, CHRISTIAN SEIGNEUR, PRAKASH KARAMCHANDANI, AND SHU-YUN CHEN

Atmospheric and Environmental Research, Inc., San Ramon, California

(Manuscript received 17 January 2006, in final form 12 April 2006)

Corresponding author address: Krish Vijayaraghavan, Atmospheric and Environmental Research, Inc., 2682 Bishop Dr., Suite 120, San Ramon, CA 94583.

E-mail: krish@aer.com

DOI: 10.1175/JAM2536.1

Copyright American Meteorological Society Sep 2007

(c) 2007 Journal of Applied Meteorology. Provided by ProQuest Information and Learning. All rights Reserved.