Atmospheric mercury chemistry: Sensitivity of global model simulations to chemical reactions

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111,, doi: /2005jd006780, 2006 Atmospheric mercury chemistry: Sensitivity of global model simulations to chemical reactions Christian Seigneur, 1 Krish Vijayaraghavan, 1 and Kristen Lohman 1 Received 19 October 2005; revised 30 March 2006; accepted 24 July 2006; published 23 November [1] The effect of new mercury (Hg) chemistry information on Hg atmospheric concentrations is investigated in a systematic manner with a global chemical transport model, taking into account current uncertainties in Hg emission and removal rates. The reactions of interest include the gas-phase oxidation of Hg(0) by O 3, the gas-phase oxidation of Hg(0) by OH, the aqueous-phase reduction of Hg(II) by HO 2 radicals, a hypothetical gas-phase reduction of Hg(II) by SO 2, and a hypothetical pseudo-first-order gas-phase reduction of Hg(II). The new kinetics of the oxidation of Hg(0) by O 3 is fast and would require balancing by a commensurate reduction reaction pathway that has not been identified; it may include some heterogeneous component and should be seen as an upper limit for atmospheric applications. Eliminating the gas-phase oxidation of Hg(0) by both O 3 and OH does not lead to realistic Hg(0) concentrations even after eliminating the aqueous-phase reduction of Hg(II) by HO 2 and having a greater dry deposition rate of Hg(0). Thus gas-phase oxidation of Hg(0) by oxidants such as O 3 and/or OH is required to reproduce global Hg(0) concentration patterns. The reduction of Hg(II) by HO 2 (or a reaction with a similar overall rate) is needed to balance the oxidation of Hg(0) by OH and O 3 but is not needed if the gas-phase oxidation of Hg(0) by OH is eliminated. The reduction of Hg(II) in power plant plumes can be represented by a reaction of Hg(II) with SO 2 ; such a reaction is consistent with the global cycling of Hg. However, a first-order reaction for Hg(II) reduction in power plant plumes is not consistent with our current understanding of the atmospheric Hg chemistry. Additional laboratory studies are recommended to address the remaining uncertainties in the atmospheric chemistry of Hg. Citation: Seigneur, C., K. Vijayaraghavan, and K. Lohman (2006), Atmospheric mercury chemistry: Sensitivity of global model simulations to chemical reactions, J. Geophys. Res., 111,, doi: /2005jd Introduction [2] Several new results on the atmospheric chemistry of mercury (Hg) have recently become available. Gårdfeldt and Jonsson [2003] challenged the reduction of Hg(II) by HO 2 that had been reported by Pehkonen and Lin [1998]. Pal and Ariya [2004a] proposed a new kinetics for the gasphase oxidation of Hg(0) by ozone (O 3 ) which is about 25 times faster than that of Hall [1995] which is currently used in most models [Ryaboshapko et al., 2002]. Calvert and Lindberg [2005] have challenged the kinetics of the gasphase oxidation of Hg(0) by O 3 and OH. Data collected in power plant plumes suggest that some Hg(II) reduction occurs in those plumes [Edgerton et al., 2003, 2004; Prestbo et al., 2004; Lohman et al., 2006]. Empirical reaction rates have been developed to represent this Hg(II) reduction using (1) a first-order Hg(II) reduction reaction [Edgerton et al., 2003, 2004; Lohman et al., 2006], or (2) a reduction of Hg(II) by SO 2 [Lohman et al., 2006]. Reactions 1 Atmospheric and Environmental Research, Inc., San Ramon, California, USA. Copyright 2006 by the American Geophysical Union /06/2005JD that occur in power plant plumes should take place elsewhere in the global atmosphere, albeit with slower kinetics because of different reactant and/or catalyst concentrations. [3] Mathematical models of the global cycling of atmospheric Hg can provide valuable information on the feasibility of such new kinetic data because they provide a framework that is constrained by the rates of Hg emissions into the atmosphere, the rates of Hg removal from the atmosphere (by dry and wet deposition) and the observed atmospheric concentrations of Hg species. Clearly, there are uncertainties with each one of these aspects of the Hg global cycle. Nevertheless, one can place some uncertainty bounds on emission rates and removal rates and the current magnitudes of Hg(0) concentrations (including their major spatial patterns) are now reasonably well established. Therefore one can use such a framework to investigate whether a specific kinetic rate is compatible with our understanding of the global cycling of Hg, taking existing uncertainties in other reaction, emission and removal rates into account. [4] We present here an analysis of the effect of these possible reactions on the global cycling of atmospheric Hg. This work focuses on annual average ground-level global Hg concentrations. Some specific areas of current research related to Arctic and Antarctic chemistry, the chemistry of the 1of17

2 Table 1. Global Anthropogenic Mercury Emission Inventory (2000 Reference Year) a Country/Continent Hg Annual Emissions, Mg/yr Reference Anthropogenic emissions 2206 United States 104 M. Houyoux, U.S. Environmental Protection Agency, Emissions data for U.S. and Canadian point and area sources, personal communication, 2004 Canada 8 M. Houyoux, U.S. Environmental Protection Agency, Emissions data for U.S. and Canadian point and area sources, personal communication, 2004 Mexico 26 Commission for Environmental Cooperation [2001] Asia 1204 Pacyna et al. [2003] Europe 239 Pacyna et al. [2003] South and Central America 92 Pacyna et al. [2003] Africa 407 Pacyna et al. [2003] Oceania 125 Pacyna et al. [2003] Natural emissions 1067 Land 625 Seigneur et al. [2004] Oceans 442 Seigneur et al. [2004] Re-emissions b 3273 Total 6546 a Boldface indicates total or subtotal. b Re-emissions are 50% of global Hg deposition [Seigneur et al., 2004]. marine boundary layer and the chemistry of the upper atmosphere are not addressed here. We use a grid-based Eulerian global chemical transport model for Hg (CTM-Hg) to simulate the global cycling of Hg. We first briefly describe the base simulation. Then, we describe our technical approach to investigate the effect of new information on Hg chemical kinetics on global Hg concentrations. Next, we present the results of the model simulations and discuss the plausibility of the various reaction kinetics in terms of global Hg cycling. Finally, recommendations are provided to reduce existing uncertainties in our understanding of the global cycling of Hg. 2. Global Mercury Base Simulation [5] The global CTM-Hg has been described by Seigneur et al. [2001, 2004]. It provides a spatial resolution of 8 latitude and 10 longitude with seven layers in the troposphere and two layers in the stratosphere. A one-year simulation is conducted using generic meteorological data obtained from a general circulation model [Hansen et al., 1983]. The global CTM-Hg is run until steady state is achieved. The inputs to the global CTM-Hg are identical to those presented by Seigneur et al. [2004] except for the anthropogenic emissions and re-emissions that were updated to represent more recent information on U.S., Canadian, and Mexican emissions and emissions from other continents. Re-emissions are currently assumed to be 50% of deposited Hg. As discussed by Seigneur et al. [2004], there is significant uncertainty with the re-emission term. Data from the METAALICUS project suggest a re-emission of only about 10% of deposited Hg(II) from land [Hintelmann et al., 2002]; on the other hand, most of deposited Hg(0) may be re-emitted. Seigneur et al. [2004] varied the ratio of current Hg emissions to preindustrial Hg emissions within a plausible range of 2 to 4 while constraining the sum of natural emissions and re-emissions to remain constant at 4268 Mg/yr. As a result, re-emissions varied between 33% and 56% of deposited Hg; there was, however, little effect on global Hg concentrations because the global amount of Hg(0) emissions remained unchanged. Table 2. Equilibria and Reactions of Atmospheric Mercury in the Base Simulation Equilibrium Process or Chemical Reaction a Equilibrium or Rate Parameter b Reference Hg(0) (g)! Hg(0) (aq) 0.11 M atm 1 Sanemasa [1975] and Clever et al. [1985] HgCl 2 (g)! HgCl 2 (aq) M atm 1 Lindqvist and Rodhe [1985] Hg(OH) 2 (g) Hg(OH) 2 (aq) M atm 1 Lindqvist and Rodhe [1985] HgCl 2 (aq)! Hg 2+ +2Cl M 2 Sillen and Martell [1964] Hg(OH) 2 (aq)! Hg 2+ +2OH M 2 Sillen and Martell [1964] Hg SO 3! HgSO 3 (aq) M 1 van Loon et al. [2001] 2 HgSO 3 (aq)+ SO 3! 2 Hg(SO 3 ) M 1 van loon et al. [2001] Hg(II) (aq)! Hg(II) (p) 34 L/g Seigneur et al. [1998] Hg(0) (g) + O 3 (g)! HgO (g) cm 3 molecule 1 s 1 Hall [1995] Hg(0) (g) + HCl(g)! HgCl 2 (g) cm 3 molecule 1 s 1 Hall and Bloom [1993] Hg(0) (g) + H 2 O 2 (g)! Hg(OH) 2 (g) cm 3 molecule 1 s 1 Tokos et al. [1998] Hg(0) (g) + Cl 2 (g)! HgCl 2 (g) cm 3 molecule 1 s 1 Ariya et al. [2002] Hg(0) (g) + OH(g)! Hg(OH) 2 (g) cm 3 molecule 1 s 1 Sommar et al. [2001] and Pal and Ariya [2004b] Hg(0) (aq) + O 3 (aq)! Hg M 1 s 1 Munthe [1992] Hg(0) (aq) + OH (aq)! Hg M 1 s 1 Lin and Pehkonen [1997] and Gårdfeldt et al. [2001] HgSO 3 (aq)! Hg(0) (aq) s 1 van Loon et al. [2000] Hg(II) (aq) + HO 2 (aq)! Hg(0) (aq) M 1 s 1 Pehkonen and Lin [1998] Hg(0) (aq) + HOCl (aq)! Hg M 1 s 1 Lin and Pehkonen [1998] Hg(0) (aq) + OCl! Hg M 1 s 1 Lin and Pehkonen [1998] a Hg(II) refers to divalent Hg species. b Parameters are for temperatures in the range of 20 to 25 C, see references for exact temperature; temperature dependence information is available for the Henry s law parameter of Hg(0) and for the kinetic rate parameter of the HgSO 3 reaction. 2of17

3 Figure 1. Annual average surface concentrations of Hg(0) from the base simulation. Table 1 presents the new anthropogenic emission inventory that represents a 2000 datum. Table 2 presents the chemical kinetic mechanism used in the base simulation. [6] The atmospheric lifetime of Hg is on the order of one year for this simulation. It must be noted, however, that Hg can cycle several times between the Hg(II) and Hg(0) species before being removed from the atmosphere (in the current model formulation, Hg(p) is assumed to be chemically inert and therefore is not involved in the Hg reductionoxidation cycle). The removal of Hg(0) from the atmosphere via dry and wet deposition is very slow because Hg(0) is assumed to have a low dry deposition velocity (here 0.01 cm/s over land on average) and is not very soluble. Assuming a planetary boundary layer height of 1000 m leads to a half-life of Hg(0) within that layer of 2.7 months via atmospheric deposition processes over land. The half-life of Hg(0) via its gas-phase reaction with OH is about 3 months (assuming 10 6 molecules of OH per cm 3 ) and that via its gas-phase reaction with O 3 is about 9 months (assuming 40 ppb of O 3 ), thereby leading to a combined half-life of 2.3 months for the gas-phase oxidation of Hg(0). Aqueous-phase oxidation further reduces the chemical halflife of Hg(0). Therefore, in the base simulation, the atmospheric half-life of Hg(0) is governed by both oxidation to Hg(II) and removal by dry deposition (Hg(0) is not removed significantly by wet deposition because of its very low solubility in water, see Table 2). [7] Removal of Hg(II) from the atmosphere occurs more rapidly than that of Hg(0) because Hg(II) species (e.g., HgCl 2, Hg(OH) 2 and HgO) are soluble and adsorb readily on most surfaces. We assumed an average dry deposition velocity of 0.5 cm/s in the global simulation, thereby leading to a half-life of 1.6 days for a 1000 m depth of the planetary boundary layer. Hg(II) has an even shorter half-life in the presence of precipitation. The reduction of Hg(II) is assumed to occur only in the aqueous phase and to be relatively rapid via its reaction with dissolved HO 2 radicals. Therefore the atmospheric half-life of Hg(II) is governed by its dry deposition in the absence of clouds, by its reaction with HO 2 radicals in the presence of nonprecipitating clouds and by wet deposition in the presence of precipitating clouds. As a result, the Hg(II) concentration field presents much stronger spatial gradients than the Hg(0) concentration field. [8] Figure 1 depicts the annual average surface concentrations of Hg(0) simulated with the global CTM-Hg. The base simulation results are similar to those presented by Seigneur et al. [2004]; minor differences in the results are due to changes in the anthropogenic emission inventory that differs by 3% on average. Those simulation results were shown to be consistent overall with available observations, particularly for Hg(0). Hg(II) and Hg(p) have shorter lifetimes than Hg(0) and their concentrations tend to reflect the influence of source areas more than those of Hg(0); consequently, the coarse spatial resolution of the global model does not capture some of the high source-related concentrations of Hg(II) and Hg(p). Accordingly, this analysis focuses on the global Hg(0) concentrations since this Hg species is more relevant to our understanding of the global Hg budget. 3. Technical Approach [9] We used the following approach to investigate in a systematic manner the effect of new information on Hg chemical kinetics on global Hg concentrations. First, we 3of17

4 Figure 2. Annual average surface concentrations of Hg(0) using the kinetics of Pal and Ariya [2004a] for the Hg(0) + O 3 reaction. investigate the effect on global Hg(0) concentrations of reactions involving the oxidation of Hg(0) to Hg(II). The corresponding simulations include using a faster kinetics for the gas-phase oxidation of Hg(0) by O 3 [Pal and Ariya, 2004a], eliminating the gas-phase oxidation of Hg(0) by O 3 [Calvert and Lindberg, 2005] and eliminating the gas-phase oxidation of Hg(0) by OH [Calvert and Lindberg, 2005]. Next, we investigate the effect on global Hg(0) concentrations of reactions that lead to the reduction of Hg(II). The corresponding simulations include eliminating the aqueousphase reduction of Hg(II) by HO 2 [Gårdfeldt and Jonsson, 2003], and introducing empirical reactions to represent the reduction of Hg(II) occurring in power plant plumes [Edgerton et al., 2003, 2004; Prestbo et al., 2004; Lohman et al., 2006]. For each simulation, we analyze whether the results lead to realistic Hg(0) concentrations and, if not, we discuss whether plausible changes in other aspects of the Hg global cycle (i.e., combined effects of changes in various reactions, Hg(0) dry deposition and emissions) could lead to realistic results. [10] Next, we conduct additional simulations to investigate combinations of changes in reactions because a change in a Hg(0) oxidation reaction may be compensated by a change in a Hg(II) reduction reaction. Finally, we address changes in emissions and Hg(0) dry deposition velocity within their plausible uncertainty range for a few cases where such changes may lead to realistic Hg(0) concentrations. 4. Gas-Phase Oxidation of Hg(0) [11] The gas-phase oxidation of Hg(0) is dominated in global models by two reactions: oxidation by O 3 and by OH (see Table 2). We investigate here the consequences of new data on the O 3 reaction kinetics and recent theoretical considerations on the kinetics and end products of the O 3 and OH reactions Faster Kinetics of the O 3 Reaction [12] The kinetics of this reaction has been evaluated by several groups. The kinetics that is currently used in most models is that of Hall [1995]. This kinetics was recently reevaluated to be about 25 times faster by Pal and Ariya [2004a]. A simulation was conducted with the kinetics of Pal and Ariya [2004a] replacing that of Hall [1995]. Figure 2 presents the corresponding surface concentrations of Hg(0). The Hg(0) concentrations are very low with values ranging from 0.3 to 1.2 ng/m 3. Such concentrations are not realistic. The new kinetics reduces the half-life for the gas-phase reaction of Hg(0) with O 3 from 9 months to about 11 days. Therefore, in this simulation this reaction governs the atmospheric half-life of Hg(0). Reducing the dry deposition velocity of Hg(0) will lead to slightly greater Hg(0) concentrations; however, it cannot lead to realistic Hg(0) concentrations (i.e., 2 to 4 times greater than those simulated here) because the half-life of Hg(0) is now governed mostly by its oxidation chemistry. [13] Several possibilities may be considered. First, the kinetics of Pal and Ariya [2004a] may be an upper bound for the reaction of Hg(0) with O 3 (e.g., because of heterogeneous processes). Second, some reduction reaction not currently in the models may exist that compensates for this fast Hg(0) oxidation; however, such a reaction has not been identified. These two possibilities should be investigated through experimental studies. Third, the global Hg emis- 4of17

5 Figure 3. Annual average surface concentrations of Hg(0) without the Hg(0) + O 3 gas-phase reaction. sions may be currently underestimated (see discussion below); a larger global emission flux could be consistent with a faster Hg(0) oxidation kinetics than currently exists in the base simulation. We investigate this hypothesis below Slower Kinetics of the O 3 Reaction [14] Calvert and Lindberg [2005] used theoretical considerations to review the mechanism and kinetics of the gasphase oxidation of Hg(0) by O 3. They concluded that the product of this reaction was a metastable HgO 3 molecule. HgO 3 would then lead to HgO formation; however, they estimated that HgO would most likely dissociate back to Hg(0) and O 3 in the gas phase. However, formation of HgO by heterogeneous reaction on surfaces could lead to the formation of stable Hg(OH) 2 in the presence of water. According to those theoretical considerations, the fast kinetics reported in laboratory experiments [e.g., Hall, 1995; Pal and Ariya, 2004a] would most likely be due to heterogeneous reactions on the laboratory vessel walls. Such heterogeneous reactions could occur in the atmosphere on aerosol particles, droplets, and ground surfaces (soil, vegetation, buildings, etc.). The surface area available in the atmosphere is likely to be significantly less than in a laboratory experiment (except perhaps in clouds and fogs), which would lead to a slower kinetics than currently used in models. Therefore it is of interest to investigate how eliminating this reaction affects Hg(0) concentrations; this approach provides a lower limit for the atmospheric kinetics of this Hg(0) + O 3 reaction. Figure 3 presents the Hg(0) surface concentrations. These concentrations range from 1.6 ng/m 3 in the Southern Hemisphere to 2.6 ng/m 3 in the Northern Hemisphere. The increase in Hg(0) concentrations with respect to the base simulation (see Figure 1) is in the range of a factor of 1.2 to 1.4; this is consistent with the change in the atmospheric half-life of Hg(0) via gas-phase reactions which increases from 2.3 to 3.1 months (the overall atmospheric half-life of Hg(0) would show a lower increase as removal rates via dry deposition and aqueousphase reactions remain unchanged). Such a change could be compensated by a decrease in emissions (although lower emissions may not be plausible), an increase in atmospheric removal (faster dry deposition of Hg(0)) and other chemical reactions (e.g., less Hg(II) reduction). A decrease in emissions may not be plausible because most uncertainties in Hg emissions tend to suggest greater emissions than used in the base simulation (see discussion below). The use of a dry deposition velocity for Hg(0) over land in the range of 0.02 to 0.03 cm/s (i.e., two to three times the value used in the base simulation) can lead to realistic Hg(0) concentrations (not shown) although the north-south gradient in Hg(0) concentrations over the oceans becomes weaker than in the base simulation because the larger landmass in the Northern Hemisphere favors Hg(0) removal in the Northern Hemisphere compared to the Southern Hemisphere. Reducing the importance of Hg(II) reduction is also feasible because of the large uncertainty that currently exists with the HO 2 reaction pathway. In summary, it appears feasible to eliminate or considerably reduce the kinetics of the gasphase oxidation of Hg(0) by O 3 because uncertainties in the dry deposition velocity of Hg(0) over land and the kinetics of the Hg(II) aqueous-phase reduction pathway can compensate for the change because of this reaction kinetics. 5of17

6 Figure 4. Annual average surface concentrations of Hg(0) without the Hg(0) + OH gas-phase reaction Slower Kinetics of the OH Reaction [15] Calvert and Lindberg [2005] used theoretical considerations to evaluate the kinetics of the gas-phase reaction of Hg(0) with OH radicals. The kinetics of this reaction has been measured in two independent laboratories using the relative rate method (OH radicals produced in the reaction vessel react with Hg(0) and another compound which has a known OH kinetics; comparing the relative rates of disappearance of Hg(0) and the other compound provides quantitative information on the kinetics of the Hg(0) + OH reaction) [Sommar et al., 2001; Pal and Ariya, 2004b]. The rates obtained were consistent within measurement error: molecule 1 cm 3 s 1 [Sommar et al., 2001] and molecule 1 cm 3 s 1 [Pal and Ariya, 2004b] at room temperature. Goodsite et al. [2004] used theoretical considerations to evaluate the kinetics and ultimate products of that reaction; they estimated a kinetics faster than measured in laboratories ( molecule 1 cm 3 s 1 at 298 K) but concluded that the reaction product HgOH was not stable because of a weak Hg-OH bond and that it would dissociate back to its precursors. Calvert and Lindberg [2005] suggest that Sommar et al. [2001] and Pal and Ariya [2004b] observed stable products because HgOH reacted with radicals (e.g., OH, HO 2 ) and molecules (e.g., NO, NO 2 ) in the laboratory to form a stable Hg(II) product. Such radicals and molecules are present at lower concentrations in the atmosphere and the formation of stable Hg(II) products would therefore be significantly limited. On the basis of theoretical considerations of Goodsite et al. [2004] and their chemical kinetic simulations, Calvert and Lindberg [2005] concluded that the Hg + OH reaction is probably unimportant in the atmosphere. This gas-phase reaction is currently the most important reaction for Hg(0) oxidation in the global Hg cycle because the occurrence of the faster Br and BrO reactions is limited in space [Ariya et al., 2004]. It is therefore of interest to investigate how its removal would affect the Hg global concentrations. [16] Bergan and Rodhe [2001] conducted global simulations with and without this reaction. They concluded that the kinetics of this reaction as measured by Sommar et al. [2001] was too fast because it led to unrealistically low Hg concentrations. However, their global model did not include aqueous-phase chemistry and, consequently, did not provide any Hg(II) reduction pathway to compensate for the fast Hg(0) oxidation. Figure 4 presents the Hg(0) surface concentrations obtained in this simulation when the Hg(0) + OH gas-phase reaction is removed. The Hg(0) concentrations are mostly above 3 ng/m 3 as the half-life of Hg(0) due to gas-phase oxidation increases from about 2.3 months to about 9 months. These concentrations are too high by about a factor of 3. The dry deposition of Hg(0) would have to be increased significantly (by about a factor of 5) to reduce those concentrations to more realistic values; moreover, the global Hg(0) concentrations do not appear realistic (not shown) because the north/south gradient that has been observed over the Atlantic ocean [Temme et al., 2003] does not appear. On the other hand, eliminating (or significantly decreasing the kinetics of) the Hg(II) reduction reactions should lead to realistic Hg(0) concentrations based on the earlier work of Bergan and Rodhe [2001]. Therefore eliminating the Hg(II) + HO 2 aqueous reaction which dominates Hg(II) reduction could account for a slower kinetics of the Hg(0) + OH reaction. 6of17

7 Figure 5. Annual average surface concentrations of Hg(0) without the Hg(0) + O 3 and Hg(0) + OH gasphase reactions Slower Kinetics of the O 3 and OH Reactions [17] If one applies the conclusions of the work by Calvert and Lindberg [2005] for both the O 3 and OH reactions, the gas-phase oxidation of Hg(0) would be considerably reduced because these two reactions dominate the chemical half-life of Hg(0) in the gas phase. To assess the effect of the lower bound scenario where these two reactions are eliminated, we conducted a simulation without these two gasphase reactions. The Hg(0) surface concentrations are presented in Figure 5. Those concentrations are too high by a factor of 3 to 4 (in the range of 5.5 to 6.4 ng/m 3 ). The other gas-phase reactions that oxidize Hg(0) to Hg(II) are slow and the aqueous oxidation of Hg(0) by O 3 and OH is significant only in clouds. Oxidation of Hg(0) by Br and BrO (not simulated here) would decrease the Hg(0) global half-life but, as mentioned above, these reactions occur in special environments and, consequently, would have limited effect on global concentrations [e.g., Ariya et al., 2004]. Decreasing the Hg(II) reduction pathway (e.g., via elimination of the Hg(II) + HO 2 reaction) and increasing the dry deposition velocity of Hg(0) (within a plausible range) may compensate to some extent the reduced oxidation rate of Hg(0); these scenarios are investigated below. 5. Reduction of Hg(II) [18] The primary pathway for Hg(II) reduction in models, the HO 2 aqueous reaction, has been challenged [Gårdfeldt and Jonsson, 2003]. In addition, a decrease in the Hg(II) fraction between the stack and a downwind measurement site has been observed in power plant plumes, which is not properly simulated by existing models. One hypothesis to explain such a change in Hg speciation is that some reaction reduces Hg(II) to Hg(0) in power plant plumes [Lohman et al., 2006]. We investigate the sensitivity of the Hg global cycle to such reactions Elimination of the HO 2 Reaction [19] Pehkonen and Lin [1998] measured the reduction of Hg(II) in aqueous solution and attributed the reaction to HO 2 radicals. This mechanism was challenged by Gårdfeldt and Jonsson [2003] who demonstrated that the intermediary monovalent species, Hg(I), would not be reduced to Hg(0) but would rather be reoxidized to Hg(II) in cloud droplets. Accordingly, Gårdfeldt and Jonsson [2003] recommended that this reaction should not be included in models of atmospheric Hg. It is therefore essential to assess the influence of this reaction on the simulated global Hg cycle. [20] A simulation was conducted without this reaction. Figure 6 presents the surface concentrations of Hg(0). These concentrations range from 0.3 to 1.2 ng/m 3 ; that is, they are significantly lower by a factor of 2 to 4 than the Hg(0) concentrations of the base simulation (see Figure 1) and, consequently, than typical observed Hg(0) concentrations. This decrease in Hg(0) concentrations is due to the fact that the major pathway for the reduction of Hg(II) to Hg(0) has been eliminated. The other Hg(II) reduction pathway is the reaction of Hg(II) with dissolved SO 2 ; it is, however, typically slow because the formation of the HgCl 2 complex is favored over that of the HgSO 3 and Hg(SO 3 ) 2 2 complexes [Seigneur et al., 1994; Pleijel and Munthe, 1995a, 1995b]. Bullock [2004] has conducted a similar numerical experiment using a continental-scale Hg model [Bullock and Brehme, 2002] and concluded that eliminating the 7of17

8 Figure 6. reaction. Annual average surface concentrations of Hg(0) without the Hg(II) + HO 2 aqueous-phase reaction of Hg(II) by HO 2 led to lower model performance for Hg concentrations in precipitation over North America. [21] Using a lower dry deposition velocity for Hg(0) would lead to greater Hg(0) concentrations. However, even assuming negligible dry deposition of Hg(0) would not lead to realistic Hg(0) concentrations because the atmospheric half-life of Hg(0) is only partially governed by deposition and its oxidation occurs sufficiently rapidly (half-life of about 2 months) to play a major role in its atmospheric halflife. Therefore eliminating Hg(0) dry deposition would lead to an increase in Hg(0) concentrations of less than a factor of 2, which would not be sufficient to lead to realistic Hg(0) concentrations. [22] Thus it appears that, for the Hg emissions considered in this global simulation, some reduction reaction of Hg(II) is required to explain the observed concentrations of Hg(0) in the atmosphere and that this reduction reaction should have a kinetics generally commensurate with that of the overall reduction of Hg(II) by HO 2 that has been used in most models. Absent this reaction, Hg emissions would need to be increased by more than a factor of 2 to lead to realistic Hg(0) concentrations, which would likely be outside the range of plausible emissions. Another possibility is that a slower kinetics of the gas-phase oxidation of Hg(0) by OH and/or O 3 (see above) could compensate for the reduced reduction of Hg(II), as explained above in the discussion of the global simulation conducted by Bergan and Rodhe [2001]. It is therefore possible to conceive a combination of plausible processes (slower Hg(0) oxidation, other Hg(II) reduction pathway, higher Hg emissions) that could compensate for the elimination of the Hg(II) + HO 2 reaction First-Order Reduction of Hg(II) [23] Ground-level measurements of speciated Hg species downwind of coal-fired power plant plumes [Edgerton et al., 2003, 2004], aircraft measurements in the plumes and measurements with a dilution chamber sampling the stack flue gases [Prestbo et al., 2004] suggest that some reduction of Hg(II) to Hg(0) takes place in coal-fired power plant plumes. Edgerton et al. [2003, 2004] estimated a Hg(II) reduction rate of 14% per hour for the identified power plant plume events; this value corresponds to a first-order kinetic rate constant of 0.15 h 1 (this value should be seen as an order-of-magnitude estimate as there were significant variations among plume events). One should note also that such a reaction is at this point hypothetical. Some Hg(II) species may be susceptible to decomposition to Hg(0) if the reaction is little endothermic (e.g., HgO [Shepler and Peterson, 2003]) whereas other Hg(II) species are considered fairly stable (e.g., HgCl 2 ). Theoretical studies of Hg speciation suggest that HgCl 2 may be the dominant Hg(II) species [Senior et al., 2000]. Nevertheless, even stable Hg(II) species can be transformed to Hg(0) according to a first-order rate, for example, via a photolytic process [e.g., Lalonde et al., 2003]. [24] A global simulation was conducted with a pseudofirst-order Hg(II) reduction reaction with a rate constant of 0.15 h 1 added (i.e., this reaction applies in this simulation to all Hg(II) in the atmosphere). The use of this empirical rate for Hg(II) reduction in the global model was intended to investigate whether a pseudo-first-order reaction with such kinetics occurring everywhere in the atmosphere (i.e., not only in power plant plumes) would be compatible with the 8of17

9 Figure 7. Annual average surface concentrations of Hg(0) with an empirical pseudo-first-order Hg(II) reduction reaction. global Hg cycle. The results are presented in Figure 7 for the Hg(0) surface concentrations. The Hg(0) concentrations are in the range of 7 to 9 ng/m 3. These concentrations are inconsistent with observations of atmospheric Hg(0) concentrations. These high concentrations result from the fast reduction of Hg(II) to Hg(0). The oxidation of Hg(0) is not sufficiently fast to compensate for the reduction of Hg(II) and, consequently, Hg(0) accumulates up to concentrations that are about 4 times greater than observations. We investigate below whether a faster oxidation of Hg(0) can compensate for this hypothetical Hg(II) reduction process Reduction of Hg(II) by SO 2 [25] Scott et al. [2003] suggested that HgO could be reduced heterogeneously by SO 2. HgCl 2 is more likely to be the predominant Hg(II) species in power plant plumes, rather than HgO [Senior et al., 2000]. Nevertheless, a reduction pathway involving HgCl 2 could be considered to occur heterogeneously in power plant plumes to explain the observed reduction of Hg(II) to Hg(0). Such a pathway is approximated here by a second-order gas-phase reaction, thereby assuming that the gas-phase concentrations of the reactants show greater variability in a plume than the particle surface area available for adsorption of the reactants. [26] Lohman et al. [2006] have used a reactive plume model [Seigneur et al., 1997] to simulate nine plume events among those identified by Edgerton et al. [2004] at the Yorkville monitoring site in Georgia, USA. Incorporating a reaction between Hg(II) and SO 2 improved model performance in terms of Hg speciation at the Yorkville site. Lohman et al. [2006] derived an empirical average second-order reaction rate of molecule 1 cm 3 s 1 and an empirical maximum rate of molecule 1 cm 3 s 1. [27] A global simulation was conducted with this reduction of Hg(II) by SO 2. The use of these empirically derived second-order rates for reduction of Hg(II) by SO 2 in the global atmosphere allows us to investigate whether such a reaction occurring everywhere in the atmosphere could be consistent with the global Hg cycle. The results are presented for the Hg(0) surface concentrations in Figure 8 using the average and maximum rates derived by Lohman et al. [2006]. These concentrations are very similar to those simulated in the base simulation. They are slightly greater than those of the base simulation because of the additional reduction of Hg(II) to Hg(0), however, the increases do not exceed 0.1 ng/m 3 with the average rate and 0.2 ng/m 3 with the maximum rate. Therefore these concentrations are consistent with observations and the reduction of Hg(II) by SO 2 in the global atmosphere (and, in particular, in power plant plumes) is compatible with our current understanding of the global atmospheric cycle of Hg. [28] Another simulation was conducted to investigate whether this Hg(II) reduction reaction could replace the reduction of Hg(II) by HO 2 to account for Hg(II) reduction in the global cycle. The empirical maximum kinetic rate was used for this simulation. The corresponding Hg(0) surface concentrations are presented in Figure 9. They are in the range of 0.3 to 1.5 ng/m 3 ; overall, they are too low by a factor of 2.5 to 4. Therefore it appears that the Hg(II) reduction by SO 2 alone is not sufficiently fast in the global atmosphere to lead to realistic Hg(0) concentrations. As mentioned above, a lower dry deposition velocity of Hg(0) 9of17

10 Figure 8. Annual average surface concentration of Hg(0) with the Hg(II) + SO 2 reaction with the (top) maximum rate of molecule 1 cm 3 s 1 and (bottom) average rate of molecule 1 cm 3 s of 17

11 Figure 9. Annual average surface concentration of Hg(0) with the Hg(II) + SO 2 reaction and without the Hg(II) + HO 2 aqueous-phase reaction. would lead to greater Hg(0) concentrations, but the corresponding increase would not be sufficient to lead to realistic Hg(0) concentrations. We investigate below whether a change in Hg emissions could compensate for the slower Hg(II) reduction kinetics. 6. Combination of Changes in Hg Oxidation and Reduction Reactions [29] A change in a Hg(0) oxidation rate can be compensated to some extent by a change in a Hg(II) reduction rate, and vice versa. It is therefore of interest to investigate those combinations that are, in theory, the most likely to lead to realistic Hg global concentrations Replacing the Hg(II) + HO 2 Reaction by the Pseudo-First-Order Hg(II) Reaction and Increasing the Hg(0) + O 3 Kinetics [30] The pseudo-first-order Hg(II) reduction at 14% per hour was shown above to lead to unrealistic Hg(0) concentrations. Minimizing the other reduction pathways of Hg(II) and increasing the oxidation kinetics of Hg(0) could help bring the simulated Hg(0) concentrations within a plausible range. As mentioned above, the reaction of Hg(II) with HO 2 is considered unlikely to occur in the atmosphere; therefore we eliminated that Hg(II) reduction reaction (note, however, that it contributes significantly less to Hg(II) reduction than the pseudo-first-order reduction reaction considered here). The reduction reaction that involves dissolved SO 2 is well established and, in any case, contributes little overall to Hg(II) reduction. As mentioned above, a faster kinetics was reported by Pal and Ariya [2004a] for the gas-phase oxidation of Hg(0) by O 3, and we incorporated this oxidation kinetics. The resulting Hg(0) surface concentrations are presented in Figure 10. They range from 2 to 4 ng/m 3 ; that is, they are still too high compared to the base simulation and observations by about a factor of 2. Increasing the Hg(0) dry deposition velocity could help reduce those concentrations. However, as mentioned above, the Hg(0) atmospheric half-life is governed in this simulation by the gas-phase oxidation of Hg(0) by O 3 (11 days versus 2.7 months for dry deposition over land). Therefore the Hg(0) dry deposition velocity would have to be increased by more than 1 order of magnitude to lead to realistic Hg(0) concentrations (i.e., >0.1 cm/s on average); such a high dry deposition velocity for Hg(0) seems unrealistic. Therefore the use of a pseudo-first-order Hg(II) reduction reaction with a rate constant of 0.15 h 1 leads to unrealistic Hg(0) concentrations whereas a second-order Hg(II) reaction with a power plant plume constituent such as SO 2 is consistent with the global Hg cycle Eliminating the Hg(0) + O 3 and Hg(0) + OH Gas-Phase Reactions and the Hg(II) + HO 2 Aqueous-Phase Reaction [31] As mentioned above, the joint elimination of the two dominant gas-phase Hg(0) oxidation reactions (i.e., reactions with O 3 and OH) leads to unrealistically high Hg(0) concentrations. On the other hand, eliminating the HO 2 aqueous reaction that governs the reduction of Hg(II) leads to unrealistically low Hg(0) concentrations. It is therefore worthwhile to investigate whether the combined elimination of these three reactions could compensate each other to such extent that they would lead to realistic Hg(0) reactions. 11 of 17

12 Figure 10. Annual average surface concentrations of Hg(0) with the empirical pseudo-first-order Hg(II) reduction reaction, without the Hg(II) + HO 2 aqueous-phase reaction and with the kinetics of Pal and Ariya [2004a] for the Hg(0) + O 3 gas-phase reaction. Figure 11. Annual average surface concentrations of Hg(0) without the Hg(0) + O 3 and Hg(0) + OH gas-phase reactions and without the Hg(II) + HO 2 aqueous-phase reactions. 12 of 17

13 Figure 12. Annual average surface concentrations of Hg(0) without the Hg(0) + O 3 and Hg(0) + OH gas-phase reactions, without the Hg(II) + HO 2 aqueous-phase reactions and with increased (4 times) Hg(0) dry deposition. Figure 11 presents the results of a simulation where those three reactions were eliminated. The Hg(0) surface concentrations range from 2.3 to 3.1 ng/m 3 ; that is, they are significantly greater (by about a factor of 2) than observed concentrations. We investigate below whether a change in the assumed Hg(0) dry deposition could account for this overestimation of Hg(0) concentrations. 7. Effect of the Hg(0) Dry Deposition Velocity [32] For cases where the chemical half-life of Hg(0) is long (a year or more), dry deposition plays a nonnegligible role in determining the atmospheric half-life of Hg(0). Increasing the dry deposition velocity of Hg(0) within plausible limits may lead to realistic Hg(0) concentrations in some cases where changing the chemical half-lives resulted in unrealistic Hg(0) concentrations. We consider here the case where the gas-phase reactions of Hg(0) with O 3 and OH and the aqueous-phase reaction of Hg(0) with HO 2 were eliminated (see above). We conducted simulations with the dry deposition velocity of Hg(0) over land increased from 0.01 cm/s to different values up to 0.1 cm/s. Such dry deposition velocities are within the range of plausible values [Lindberg and Stratton, 1998; Xu et al., 1999; Poissant et al., 2004; Lindberg et al., 2004]. The results of the simulation conducted with a dry deposition velocity over land of 0.04 cm/s are presented for annual Hg(0) surface concentrations in Figure 12. These results suggest that slow oxidation of Hg(0) (through the aqueousphase reactions with O 3 and OH) without significant Hg(II) reduction may lead to realistic global concentrations (range of 1.4 to 2.2 ng/m 3 ) if one assumes a sufficiently high, yet plausible, Hg(0) dry deposition velocity. However, the north/south gradient of Hg(0) concentrations observed over the Atlantic ocean [Temme et al., 2003] is not reproduced here because the faster Hg(0) dry deposition process occurs only over land (one assumes that there is no Hg(0) deposition over the oceans because of its very low solubility) and affects Hg(0) concentrations more in the Northern Hemisphere than in the Southern Hemisphere because of a greater landmass in the former. Therefore the results of this simulation are not realistic and they suggest that some gas-phase oxidation of Hg(0) by oxidants such as O 3 and OH is needed to reproduce the observed spatial patterns of Hg(0) concentrations. 8. Effect of Mercury Emissions on Global Mercury Concentrations [33] The global Hg simulations presented above used the global emission inventory of Seigneur et al. [2004] with some modifications reflecting updates for the anthropogenic emissions. There are, however, considerable uncertainties with current estimates of global emissions. For example, anthropogenic emissions from Asia could be underestimated by a factor of 2 [Jaffe et al., 2005], emissions of mobile sources that are typically not included in emission inventories could be significant based on measurements conducted in the United States [Edgerton and Jansen, 2004], emissions from volcanoes could be underestimated by as much as a factor of 5 [Pyle and Mather, 2003], other natural emissions could also be underestimated [Lindberg et 13 of 17

14 Figure 13. Annual average surface concentration of Hg(0) using the kinetics of Pal and Ariya [2004a] for the Hg(0) + O 3 reaction and with global emissions twice those of the base simulation. al., 2004], and the paucity of data on the fraction of deposited Hg that is reemitted [Hintelmann et al., 2002] leads to some large uncertainty on this possibly important atmospheric input [Seigneur et al., 2004]. Consequently, it is of interest to investigate how the uncertainties in Hg emissions may affect the results obtained above with the various atmospheric chemical mechanisms. Since Hg emissions are more likely to be underestimated than overestimated, this sensitivity analysis focuses on the results that showed underestimations of the Hg(0) concentrations. [34] We investigated two cases where increased emissions may compensate for low Hg(0) concentrations obtained in a sensitivity simulation. First, we considered the faster gasphase O 3 kinetics because this simulation led to Hg(0) concentrations that were too low to be realistic. All base emissions (anthropogenic, natural and re-emissions) were doubled. We did not attempt to spatially distribute the increase in emissions. Natural emissions, re-emissions and a significant fraction of anthropogenic emissions consist of Hg(0) which has a long atmospheric half-life, and therefore such an across-the-board treatment of the incremental emissions should not have a significant effect on annual Hg(0) surface concentrations and it will serve our purpose here. The simulation results are presented in Figure 13. The Hg(0) concentrations are below 1.3 ng/m 3 except for an area in East Asia where they reach 2.2 ng/m 3. In addition, there is no apparent north/south gradient in the Hg(0) concentrations over the oceans. Such a gradient has been reported over the Atlantic ocean by Temme et al. [2003] and is reproduced in the base simulation (see Figure 1). The reason for this lack of a gradient is that Hg(0) is depleted too fast in the Northern Hemisphere. Therefore these results are not realistic and they suggest that the kinetics of Pal and Ariya [2004a] for the ozone reaction may represent both homogeneous and heterogeneous processes and should be seen as an upper limit for atmospheric conditions. [35] Next, the global simulation with no reduction of Hg(II) by HO 2 but reduction of Hg(II) with SO 2 was conducted with the same doubling of Hg emissions across the board. The results of this simulation are presented in Figure 14. The Hg(0) surface concentrations are in the range of 0.7 to 2.5 ng/m 3. The Hg(0) concentrations are slightly lower than those of the base simulation and it appears that the longitudinal gradients in Hg(0) concentrations are stronger than in the base simulation (see Figure 1) because of the slower reduction of Hg(II). Nevertheless, these results indicate that a chemical mechanism that seems to lead to unrealistic Hg concentrations using our current global emission inventory could lead to more realistic Hg concentrations using a different, yet plausible emission inventory. Therefore the existing uncertainties in the global emissions of Hg have a significant effect on our current characterization of global Hg concentrations and they must be addressed in order to better constrain the global Hg cycling budget. 9. Conclusion [36] Several recent results on Hg chemical transformations were investigated with a global chemical transport model for Hg. These reactions included the gas-phase oxidation of Hg(0) by O 3, the gas-phase oxidation of Hg(0) by OH, the aqueous-phase reduction of Hg(II) by HO 2 radicals, a pseudo-first-order gas-phase reduction of Hg(II) and the gas-phase reduction of Hg(II) by SO 2. Table 3 14 of 17

15 Figure 14. Annual average surface concentration of Hg(0) with the Hg(II) + SO 2 reaction and without the Hg(II) + HO 2 aqueous-phase reaction, using global emissions twice those of the base simulation. Table 3. Summary of Sensitivity Simulations Simulation Range of [Hg(0)], ng/m 3 Dry Deposition Possible Adjustments Emissions Other Reactions Conclusion 1. Faster O 3 (g) reaction no a see 12 see 9 not realistic, change needed in emissions or Hg(II) reduction 2. No O 3 (g) reaction yes b no c yes b plausible with a change in Hg(0) dry deposition and/or Hg(II) reduction 3. No OH(g) reaction no b,d no c yes e plausible with a decrease in Hg(II) reduction 4. No O 3 (g) and OH(g) reactions no f no c see 10 not realistic, change needed in Hg(II) reduction 5. No HO 2 (aq) reaction no a no g yes, e see also 8 plausible with a decrease in Hg(0) oxidation 6. Pseudo-first-order Hg(II) reduction 7 9 no f no c see 9 not realistic, change needed in other reactions 7. SO 2 (g) reaction NA h NA h NA h plausible 8. SO 2 (g) reaction and no HO 2 (aq) reaction no a see 13 NA h not realistic, change needed in emissions 9. Pseudo-first-order Hg(II) reduction, 2 4 no f no c no i not realistic faster O 3 (g) reaction and no HO 2 (aq) reaction 10. No O 3 (g), OH(g) and HO 2 (aq) reactions see 11 no c NA h not realistic, change needed in Hg(0) dry deposition 11. Same as 10 with Hg(0) dry deposition NA h no c NA h not realistic, no north/south gradient 12. Faster O 3 (g) reaction with emissions no a NA h no i not realistic, no north/south gradient 13. Same as 8 with emissions no a NA h NA h plausible a Reducing Hg(0) dry deposition would have little effect because chemistry governs the Hg(0) atmospheric lifetime in this simulation. b Simulation results are not shown here. c Emissions would need to be reduced significantly which is not realistic. d Global spatial patterns of Hg(0) concentrations are not realistic. e See Bergan and Rodhe [2001] for example. f Large increase in Hg(0) dry deposition would be required (> 10) which is not realistic. g Large increase in Hg emissions would be required (> 2) which is not realistic. h NA, not applicable, either because the chemical mechanism already leads to plausible Hg(0) concentrations or because adjustments in those processes have already been made. i No known reaction pathways could compensate. 15 of 17

16 presents a summary of the sensitivity simulations in this study. The global simulation results obtained with the current assumptions made regarding the emissions and removal rates of Hg species suggest the following. [37] The new kinetics of the oxidation of Hg(0) by O 3 [Pal and Ariya, 2004a] is fast and would require a commensurate but unidentified reduction reaction to lead to realistic Hg concentrations. An increase in Hg emissions by a factor of 2 or 3 (i.e., within a plausible range of uncertainty) does not lead to realistic Hg(0) concentrations because the north/south Hg(0) concentration gradient that has been observed over the Atlantic ocean is not reproduced. This kinetics may include both homogeneous and heterogeneous processes and could therefore be considered as an upper limit under atmospheric conditions. [38] A reduction reaction with an overall rate similar to that of the reduction of Hg(II) by HO 2 is needed to balance the oxidation of Hg(0) by OH and O 3 currently used in models. However, if the gas-phase oxidation of Hg(0) by O 3 and OH is lower than currently assumed in models, as suggested by theoretical considerations, then, the kinetics of the reduction of Hg(II) should be decreased accordingly. For example, the reduction of Hg(II) by HO 2 is not needed if the gas-phase oxidation of Hg(0) by OH is eliminated. However, eliminating the Hg(0) gas-phase oxidation by both O 3 and OH (or assuming that they both have a negligible contribution to Hg(0) oxidation overall) does not lead to realistic results. One can obtain Hg(0) concentrations that appear to have realistic levels overall by jointly eliminating the Hg(II) + HO 2 reaction and increasing the Hg(0) dry deposition velocity (within its plausible range of uncertainty); however, the simulation results do not reproduce the north/south Hg(0) gradient observed over the Atlantic ocean. Therefore some gas-phase oxidation of Hg(0) by oxidants such as O 3 and OH is needed to obtain realistic Hg(0) concentrations. [39] The proposed reduction of Hg(II) in power plant plumes can be represented by a reaction of Hg(II) with SO 2 ; this reaction does not affect the global cycling of Hg significantly. However, a pseudo-first-order Hg(II) reduction reaction derived from power plant plume data is not consistent with our current understanding of the atmospheric Hg chemistry because its kinetics would be too fast and would lead to Hg(0) concentrations that are too high. [40] The kinetics of the OH and O 3 gas-phase reactions should continue to be investigated as they are potentially key reactions for Hg(0) oxidation. We also recommend that reactions that could reduce Hg(II) continue to be investigated. The reduction of Hg(II) by SO 2 could be incorporated as a surrogate reaction to represent Hg(II) reduction in power plant plumes until a more definitive mechanism is identified in the laboratory. [41] Uncertainties in the current emissions and removal rates of Hg species are still significant and, as a result the global cycling of Hg is poorly constrained. For example, we showed that replacing the Hg(II) reduction by HO 2 by the Hg(II) reduction with SO 2 could lead to realistic results if Hg emissions are doubled. These uncertainties need to be reduced before we can ensure that the chemical kinetic mechanism used in the chemical transport models properly represents the global Hg cycle. [42] Acknowledgments. This work was funded by EPRI, Palo Alto, California, USA, under contract EP-P15776/C7851. Thanks are due to the EPRI Project Manager, Leonard Levin, for his continuous support and constructive comments. References Ariya, P. A., A. Khalizov, and A. Gidas (2002), Reactions of gaseous mercury with atomic and molecular halogens: Kinetics, product studies, and atmospheric implications, J. Phys. Chem., 106, Ariya, P. A., et al. (2004), The Artic: A sink for mercury, Tellus, Ser. B, 56, Bergan, T., and H. Rodhe (2001), Oxidation of elemental mercury in the atmosphere: Constraints imposed by global scale modeling, J. Atmos. Chem., 40, Bullock, O. R., Jr. (2004), Aqueous reduction of Hg 2+ to Hg 0 by HO 2 in the CMAQ-Hg model, RMZ Mater. Geoenviron., 51, 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, Calvert, J. G., and S. E. Lindberg (2005), Mechanisms of mercury removal by O 3 and OH in the atmosphere, Atmos. Environ., 39, Clever, H., S. A. Johnson, and E. M. Derrick (1985), The solubility of mercury and some sparingly soluble mercury salts in water and aqueous solutions, J. Phys. Chem. Ref. Data, 14, Commission for Environmental Cooperation (2001), Preliminary atmospheric emissions inventory of mercury in Mexico, Rep , Montreal, Que., Canada. Edgerton, E. S., and J. J. Jansen (2004), Elemental Hg measurements in Atlanta, GA, USA: Evidence for mobile sources?, RMZ Mater. Geoenviron., 51, Edgerton, E., J. Jansen, and B. Hartsell (2003), Field observations of mercury at a rural/urban pair of sites near Atlanta, Georgia, USA, paper presented at International Conference on Air Quality IV, Energy and Environ. Res. Cent., Univ. of N. D., Arlington, Va., Sept. Edgerton, E. S., J. J. Jansen, and B. E. Hartsell (2004), Speciated Hg measurements at a rural site near Atlanta, GA, USA, RMZ Mater. Geoenviron., 51, Gårdfeldt, K., and M. Jonsson (2003), Is bimolecular reduction of Hg (II)- complexes possible in aqueous systems of environmental importance?, J. Phys. Chem., 107, Gårdfeldt, K., J. Sommar, D. Stroemberg, and X. Feng (2001), Oxidation of atomic mercury by hydroxyl radicals and photoinduced decomposition of methylmercury in the aqueous phase, Atmos. Environ., 35, Goodsite, M. E., J. M. C. Plane, and H. Skov (2004), A theoretical study of the oxidation of Hg 0 to HgBr 2 in the troposphere, Environ. Sci. Technol., 38, Hall, B. (1995), The gas-phase oxidation of elemental mercury by ozone, Water Air Soil Pollut., 80, Hall, B., and N. Bloom (1993), Investigation of the chemical kinetics of mercury reactions, report, EPRI, Palo Alto, Calif. Hansen, J., G. Russel, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R. Ruedy, and I. Travis (1983), Efficient three-dimensional global models for climate studies: Models I and II, Mon. Weather Rev., 111, Hintelmann, H., R. Harris, A. Heyes, J. P. Hurley, C. A. Kelly, D. P. Krabbenhoft, S. Lindberg, J. W. M. Rudd, K. J. Scott, and V. L. St. Louis (2002), Reactivity and mobility of new and old mercury deposition in a boreal forest ecosystem during the first year of the METAALICUS study, Environ. Sci. Technol., 36, Jaffe, D., E. Prestbo, P. Swartzendruber, P. Weiss-Penzias, S. Kato, A. Takami, S. Hatakeyama, and Y. Kajii (2005), Export of atmospheric mercury in Asia, Atmos. Environ., 39, Lalonde, J. D., M. Amyot, M. Doyon, and J. Auclair (2003), Photo-induced Hg(II) reduction in snow from the remote and temperate Experimental Lakes Area (Ontario, Canada), J. Geophys. Res., 108(D6), 4200, doi: /2001jd Lin, C. J., and S. O. Pehkonen (1997), Aqueous-free radical chemistry of mercury in the presence of iron oxides and ambient aerosol, Atmos. Environ., 31, Lin, C. J., and S. O. Pehkonen (1998), Oxidation of elemental mercury by aqueous chlorine (HOCl/OCl ): Implications for tropospheric mercury chemistry, J. Geophys. Res., 103, 28,093 28,102. Lindberg, S. E., and W. J. Stratton (1998), Atmospheric mercury speciation: Concentrations and behavior of reactive gaseous mercury in ambient air, Environ. Sci. Technol., 32, Lindberg, S. E., D. Porcella, E. Prestbo, H. Friedli, and L. Radke (2004), The problem with mercury: Too many sources, not enough sinks, RMZ Mater. Geoenviron., 51, Lindqvist, O., and H. Rodhe (1985), Atmospheric mercury A review, Tellus, Ser. B, 37, of 17