Interactive comment on “ Investigation of processes controlling GEM oxidation at mid-latitudinal marine , coastal , and inland sites ”

This paper uses a box model to study the controlling processes of GEM oxidation (or GOM formation) at different types of surface sites, and provides new and important information on the chemistry mechanisms of mercury that might occur in the real atmosphere. It fits well into the scope of ACP. I recommend the paper for publication after addressing the following comments. A major comment is that the box model simulation results should be compared against the measurements of PBM mixing ratios at these sites. This would help the interpretation of some controlling processes such as gas-particle partitioning in the model. Another general comment is that a more detailed description of the box model set up should be given in the paper, for example the exchange of GOM between the free troposphere and the boundary layer. A schematic


Introduction
Mercury (Hg) is a toxic pollutant found globally in air, natural waters, and soils.The health concern of Hg arises from the neurotoxic organic form, methyl mercury (MeHg), in the aquatic environments (Mason et al., 2006;Miller et al., 2007;Rolfhus et al., 2003).The high bioaccumulation and biomagnification of MeHg lead to human exposure through the consumption of seafood (Clarkson, 1994) In the atmosphere, Hg exists in three forms: gaseous elemental mercury (GEM), gaseous oxidized mercury (GOM), and particulate bound mercury (PBM).The majority of atmospheric Hg is GEM, comprising > 95 % of total gaseous mercury (TGM = GEM + GOM).The 0.8-1.7 years atmospheric lifetime of GEM is conducive to long range transport of Hg as a global pollutant (Bergan et al., 1999;Bergan and Rodhe, 2001;Holmes et al., 2006;Lin and Pehkonen, 1999;Schroeder and Munthe, 1998;Selin et al., 2007).In contrast, GOM and PBM are relatively short-lived and subject to dry and wet deposition due to their high solubility in water and low vapor pressure.GOM in the atmosphere can be produced from oxidation of GEM, released directly from anthropogenic emissions, and transformed from PBM. Oxidation of GEM was usually thought to be a major source of GOM in remote regions.
Chemical speciation of atmospheric Hg is essential to understand its geochemical cycle.Theoretical and experimental studies suggested that the main oxidants of GEM in the atmosphere are ozone (O 3 ), hydroxyl radical (OH), atomic bromine (Br), bromine monoxide (BrO), hydrogen peroxide (H 2 O 2 ), and atomic chlorine (Cl), yielding GOM species of HgO, HgBrO, HgBr, Hg(OH) 2 , HgCl, and through further reaction to other mercury halides (Ariya et al., 2015;Dibble et al., 2012;Lin and Pehkonen, 1999).Although efforts have been made to investigate the relative importance of these oxidants for GEM oxidation in the troposphere, it is still not well understood.In the terrestrial environment, it was suggested that the oxidation of GEM was primarily by O 3 and OH radicals (Shon et al., 2005;Sillman et al., 2007).In recent years, a consensus has emerged that the GEM + O 3 reaction most likely occurs with solid-phase products, whose speciation and quantification remain unknown (Ariya et al., 2015;Pal and Ariya, 2004b;Rutter et al., 2012;Snider et al., 2008).The reaction of GEM+OH has been subject to debate between theoretical and experimental studies, as no mechanism that is consistent with thermochemistry has been proposed (Ariya et al., 2015;Pal and Ariya, 2004a;Subir et al., 2011).In the MBL, measurements of GOM in the polar regions (Simpson et al., 2007;Steffen et al., 2008) to sub-tropical MBL (Laurier et al., 2003;Laurier and Mason, 2007;Obrist et al., 2011) (Holmes et al., 2009(Holmes et al., , 2010;;Kim et al., 2010;Lindberg et al., 2002;Obrist et al., 2011;Soerensen et al., 2010;Toyota et al., 2014;Wang et al., 2014;Xie et al., 2008) have also suggested Br as an important oxidant of GEM.The major source of atmospheric Br was suggested to be produced photolytically from Br-containing compounds and through the Br / BrO cycle involving tropospheric O 3 (Saiz-Lopez and Glasow, 2012;Simpson et al., 2015).Hg chemistry in the MBL, the lowest part of the troposphere in direct contact with the sea surface, has global importance as approximately 70 % of the earth's surface is covered by oceans (Glasow et al., 2002).Hg in the MBL cycles differently from in coastal or inland areas.However, contemporary models are not able to reproduce GOM observations temporally and spatially due to knowledge gaps in Hg science, simplified model assumptions, and uncertainties of measurements (Ariya et al., 2015;Lin et al., 2006).In the polar region, bromine radicals were identified as the primary cause of the Arctic mercury depletion events (AMDE) (Kim et al., 2010;Lindberg et al., 2002;Toyota et al., 2014;Xie et al., 2008).In the MBL outside Polar Regions, due to lower mixing ratios of atmospheric halogen radicals, often lower than the detection limit, mechanisms for GOM production were more controversial than the Polar Regions.Using a box model, Hedgecock et al. (Hedgecock et al., 2003;Hedgecock andPirrone, 2004, 2005) suggested that O 3 was a dominant GEM oxidant in the MBL at mid-latitudes in Mediterranean region, and that the GEM+O 3 reaction may form solid products.However, the reaction kinetics in their model were out-of-date with limited halogen chemistry, and fixed emission used in the model oversimplified the source terms.Holmes et al. (2009) simulated that GEM oxidation by Br comprised 35-60 % of the GOM sources using BrO concentrations calculated at a photostationary state from a prescribed distribution of Br mixing ratios.In this study a parameter was introduced to account for entrainment of free tropospheric GOM into the MBL and the Br mixing ratio was adjusted to capture the observed GOM diurnal trend, which could cause large uncertainties in GOM simulations.Most recently, Wang et al. (2014) employed updated Hg reactions together with bromine and iodine reactions, adopting the free tropospheric GOM entrainment Introduction

Conclusions References
Tables Figures

Back Close
Full parameter from Holmes et al. (2009) for tropical MBL, and found Br to be a primary GEM oxidant, but oxidation by Br or O 3 / OH alone was unable to reproduce observed GOM concentration.
In this study, we employed a state-of-the-art chemical mechanism that incorporates gas and aqueous phase chemistry of Hg, O 3 , and halogen to investigate the dynamics of GOM formation under various atmospheric conditions in mid-latitude regions.The most up-to-date kinetics was applied.Halogen radical mixing ratios (such as Br and BrO) were calculated using up-to-date atmospheric halogen reactions.Clear sky days with calm wind conditions were selected to minimize the entrainment effect of free tropospheric air and regional transport.Moreover, the initial GEM mixing ratios in the model were obtained from observations in three different environments and were set to be constant mimicking GEM emission flux.

Box model description
The Kinetic PreProcessor version 2.1 (Sandu and Sander, 2006) was utilized as the framework of the box model (Hedgecock et al., 2003;Hedgecock andPirrone, 2004, 2005).A second order Rosenbrock method (Verwer et al., 1999) was applied to solve the coupled ordinary differential equations.The box model used in this study was initially set up by Kim et al. (2010).It was further improved in this study by incorporating the most up-to-date gas and aqueous phase chemical mechanisms (Atkinson et al., 2004;Dibble et al., 2012;Sander et al., 2011) to the model.

Reactions and kinetics
The box model has a total of 424 reactions: 276 gas-phase reactions (including Hg, halogen, O 3 , sulfate, and hydrocarbon reactions), 52 gas-water equilibriums, 28 aqueous equilibriums and 68 aqueous reactions.Most of these reactions and kinetic data Introduction

Conclusions References
Tables Figures

Back Close
Full were updated based on JPL Report No. 17 (Sander et al., 2011), Atkinson et al. (2004), and the references listed in Table 1.Photodissociation coefficients were calculated from the Tropospheric Ultraviolet and Visible (TUV) Radiation Model (Madronich, 1993).
Aqueous Hg reactions include: or replaced by simple approximation as previous Hg box model studies outside of Polar regions did.Hence, the most up-to-date halogen chemistry from the literature was included in our model.

Initial conditions and input data
Observations at three sites from the University of New Hampshire (UNH) AIRMAP Observing Network (http://www.eos.unh.edu/observatories/data.shtml) were used: a marine site located on Appledore Island (AI) in the Gulf of Maine (42.97  et al. (2009, 2012) and boundary layer height estimated from Mao and Talbot (2004).
Other physical parameters (i.e.Henry's constants, liquid water content, and aerosol radius) were used to simulate the gas-particle partitioning process in the box model.Introduction

Conclusions References
Tables Figures

Back Close
Full

Gas-particle partitioning
An empirical expression was utilized to calculate particle size growth relative to its dry radius (r dry ) (Lewis and Schwartz, 2006): where RH is the relative humidity, and r is the particle radius at RH.
Gas-particle partitioning was treated by mass transfer between droplets and air.The dynamic mass transfer coefficient across the gas-aqueous interface was calculated using the method developed by Schwartz (1986).The net mass flux (F , molecule cm −3 s −1 ) between the gas and aqueous phase is given by where L is the liquid water content (m 3 water m −3 air ), k mt is the mass transfer coefficient (s −1 ), c g is the gas phase concentration of the species (molecules cm −3 ), c aq is the aqueous phase concentration of species (molecules cm −3 ), H is the Henry's constant of the species (M atm −1 ), R is the universal gas constant (atm L K −1 mol −1 ), and T is atmospheric temperature (K).k mt is calculated as follow: where r is the particle radius (µm), D g is the diffusion coefficient (m 2 s −1 ), v is the mean thermal molecular velocity (m s −1 ), α is the dimensionless accommodation coefficient, and M is the species molecular weight (g mol −1 ).Introduction

Conclusions References
Tables Figures

Back Close
Full

Case selection
A total of 83 cases were examined to investigate the role of chemistry in Hg cycling in the MBL, coastal, and inland environments.At the study sites, significant warm season declines of GEM were observed with annual maximums in spring and minimums in autumn resulting in seasonal amplitudes up to 100 ppqv at TF (Mao et al., 2008).The lost GEM during the warm season most likely entered the ecosystem.Chemical transformation of GEM in warm seasons was suspected to be one of the factors causing the observed seasonal decline in GEM.As such, this study selected the cases representing summer days when chemical processes were most likely dominant.To exclude the influence of wet deposition, we selected clear-sky conditions based on the observed photodissociation rate constant of NO 2 (j NO 2 ) and solar radiation flux.To minimize the influence of transport, cases with arithmetic daily mean wind speed higher than 75 % percentile of all summer days in studied years (> 1.3 m s −1 at TF, > 6 m s −1 at PM, and > 7 m s −1 at AI) were excluded.As a result, 50, 12, and 21 clear-sky days at AI (marine), TF (coastal), and PM (inland, elevated), respectively, were selected from summers of 2007, 2008, and 2010.Since there was no temperature data available for summer 2009 at TF, 2009 was not considered.

Backward trajectory model
The

Model evaluation
To evaluate the box model performance with observations, the following statistical performance measures (Chang and Hanna, 2004;Hanna, 1988;Hanna et al., 1991Hanna et al., , 1993)), which include the functional bias (FB), the normalized mean square error (NMSE), the root mean square error (RMSE), and the partition of NMSE due to systematic errors (NMSE s ) were used: where C p is model predictions, C 0 is observations, overbar (C) is average over the dataset.

General characteristics in measured GOM and GEM
In the selected 83 cases, atmospheric GOM and GEM mixing ratios varied greatly at the three sites (Fig. 2).Mixing ratios of GOM varied over 0.03-87.79ppqv at AI, 0.04-4.93ppqv at TF, and 0-0.65 ppqv at PM. GOM did not show consistent diurnal variation at these sites.At AI and TF, significant diurnal variation was observed with afternoon maximums and nighttime minimums.At AI, GOM peaked at 10 ppqv over 14:00-16:00 EDT and was ∼ 5 ppqv at night, well above the limit of detection (LOD, ∼ 0.1 ppqv, from Sigler et al., 2009).At TF, GOM mixing ratios peaked at 0.75 ppqv at Introduction

Conclusions References
Tables Figures

Back Close
Full 17:00 EDT and were below LOD at night, before 08:00 EDT.The GOM diurnal cycle at PM was different from that at AI and TF.At PM, averaged GOM had higher mixing ratios at night and in the early morning than in the afternoon.However, the median values were showing afternoon peaks and nighttime minimums.The difference between average and median GOM diurnal cycles was driven by 3 cases that had abnormally high GOM mixing ratios (> 0.6 ppqv) at night or in the early morning relative to the average GOM mixing ratio through the day (∼ 0.1 ppqv).
Mixing ratios of GEM ranged over 65-231 ppqv at AI, 60-213 ppqv at TF, and 121-231 ppqv at PM (Fig. 2).On average, GEM mixing ratios at PM were 8 % higher than that at TF and 12 % higher than that at AI. Unlike GOM, GEM diurnal cycles showed nearly flat patterns at AI and PM, though slightly higher (∼ 3 %) GEM mixing ratios at night than in the daytime were observed at PM.In contrast, the average GEM diurnal cycle at TF showed an early morning (07:00 EDT) minimum (112 ppqv) and a daytime (13:00 EDT) maximum (153 ppqv).
The site differences of GOM and GEM diurnal cycles could be attributed to different chemical environments, land surface types, and meteorological conditions.For example, the GEM daily minimum at night and in the early morning at TF was likely caused by a strong net loss dominated by dry deposition under nocturnal inversion (Mao et al., 2008;Mao and Talbot, 2012).Nocturnal inversion also influenced the GEM and GOM diurnal cycles at PM, albeit differently from at TF.The elevation of PM site is 700 m a.s.l., above the nocturnal inversion layer (< 200 m) (e.g.Kutsher et al., 2012), and thus GEM and GOM at night were continuously replenished by those produced from daytime and remaining in the residual layer, which likely caused higher nighttime values at PM. Daytime peaks of GOM at TF and AI were most likely caused by photochemical oxidation of GEM under strong solar radiation.The causes for such variation were examined in Sect.3.2.2.Introduction

Conclusions References
Tables Figures

Back Close
Full

Simulated diurnal variation and speciation of GOM
Model simulated diurnal cycles of GOM averaged over the 50, 12, and 21 clear-sky days at AI, TF, and PM, respectively, were shown in Fig. 3.The patterns of diurnal variation were similar at the three sites with daily peaks at ∼ 14:00 LT, but the magnitude varied by site.AI had the largest GOM diurnal amplitude (i.e., daily maximum -daily minimum) ranging from 0.45 to 20.99 ppqv, TF from 0.02 to 1.85 ppqv, and PM showed negligible diurnal variation.Similar magnitude variation was also exhibited in GOM observations (Fig. 2).Overall, simulated GOM mixing ratios at the three sites were in agreement with observations (detailed comparison in Sect.3.4).
The simulations suggested that the dominant GOM species and GEM oxidants varied by site (Fig. 4).At AI, brominated GOM species comprised 50-71 % of the total GOM over 09:00-16:00 EDT, whereas HgO was dominant (56-92 % of the total GOM) during the remaining day.At TF and PM, HgO was the predominant GOM species (80-99 %).HgO was produced from oxidation of GEM by O 3 and OH.The contribution to HgO from oxidation by O 3 was larger than by OH except at noon when OH mixing ratios reach daily peaks resulting in comparable contributions (48 and 52 % by OH and O 3 , respectively).At AI, HgBrO, BrHgOOH, and BrHgOBr were the most abundant brominated GOM species, which constituted ∼ 99 % of the total brominated GOM.Hg-BrO was produced from the GEM + BrO reaction, while BrHgOOH and BrHgOBr were produced from GEM oxidation by Br radicals followed by reactions of HgBr with HO 2 and BrO.Hg(OH) 2 from GEM oxidation by H 2 O 2 appeared to be an important nighttime GOM species at the inland site (PM), accounting for 33 % of the total GOM at night.Other GOM species were negligible in the studied cases.Introduction

Conclusions References
Tables Figures

Back Close
Full

Sensitivity of GOM to physical and chemical parameters
The base case (Case 0) of these sensitivity runs represented the real atmospheric conditions on the selected 50 days at AI. Case 1-10 are sensitivity cases where one parameter in the base case was changed (Table 3).Case 1 turned off photolysis reactions.
Cases 2-4 tested the gas-particle partitioning scheme.Cases 5-8 tested the sensitivity of GOM mixing ratios to GEM oxidation reactions and their coefficients.Cases 9-10 tested the sensitivity of GOM mixing ratios to temperature.The importance of photochemical radicals in GEM oxidation was demonstrated clearly in decreases of 21-80 and 28-92 % in daytime GOM and PBM, respectively with largest decreases at noon as a result of turning off photochemistry (Case 1).Case 2 showed ∼ 74 % of oxidized Hg transformed to PBM at AI with gas-particle partitioning switched on.In this case, HgO and Hg(OH) 2 were more sensitive than halogenated GOM species (such as BrHgOOH and BrHgOBr).Turning off gas-particle partitioning more than quadrupled the mixing ratios of HgO and Hg(OH) 2 throughout the day compared to increases of more than 100 and 60 % halogenated GOM species during daytime and nighttime, respectively.
Decreasing liquid water content by 1 order of magnitude tripled GOM mixing ratios, whereas increasing the same amount decreased GOM by 87 % (Cases 3-4).Sensitivity of GOM and PBM mixing ratios to dominant GEM oxidation reactions are shown in 5 %, respectively, since Br and OH are both photochemical radicals and O 3 was the predominant oxidant for GEM in the model.Cases 9-10 suggested that nighttime GOM and PBM mixing ratios were more sensitive to temperature than those during daytime.Increasing temperature by 10 K caused a 9 % increase each in GOM and PBM mixing ratios during daytime but a decrease of 13 % in GOM and 54 % in PBM at night.This was because the rate coefficient of GEM + O 3 increases with increasing temperature, but the rate coefficient of GEM + OH decreases with increasing temperature.

Influence of physical and chemical processes on GOM diurnal cycle
Large variations were exhibited in both observed (Fig. 2) and simulated (Fig. 3) GOM mixing ratios at AI, TF, and PM.Considering that all cases were under relatively calm, clear-sky conditions, the simulated GOM mixing ratio and diurnal cycle were controlled primarily by chemical reactions, dry deposition, and gas-particle partitioning.To quantify the contribution of processes to the difference of GOM mixing ratios at the three sites, three sensitivity cases were conducted at TF and PM: (1) use the same dry deposition velocity and boundary layer height as those of AI for TF (denoted as TF_AIdry) and PM (denoted as PM_AIdry), (2) use the same gas-particle partitioning parameters as those of AI for TF (denoted as TF_AIaero) and PM (denoted as PM_AIaero), (3) use the same physical parameters as those of AI for TF (denoted as TF_AIaerodry) and PM (denoted as PM_AIaerodry).
Comparison of simulated GOM diurnal cycles from the AI, TF_AIaerodry and PM_AIaerodry cases showed the influence of different chemical scenarios on GOM mixing ratios at the three sites.At night, GOM mixing ratios at the three sites did not vary significantly (0-2 ppqv), with higher values at PM than those at AI and TF (Fig. 5).However, the mid-day peak at AI was more than one order of magnitude higher than those in the PM_AIaerodry and TF_AIaerodry cases, indicating more chemical transformation of Hg occurring at AI.The daytime mixing ratios of GOM at TF and PM were similar, while the nighttime GOM mixing ratios at PM were 30-52 % higher than at AI Introduction

Conclusions References
Tables Figures

Back Close
Full and 20-200 % higher than at TF.This probably resulted from larger nighttime GEM and O 3 mixing ratios, hence producing more GOM, at PM than those at TF and AI.Specifically, nighttime GEM mixing ratios at PM were 8-15 % higher than at AI and 8-34 % higher than TF cases, while nighttime O 3 mixing ratios at PM were 11-70 % larger than at AI and 35-260 % larger than at TF. PM had higher nighttime GEM and O 3 mixing ratios, because this site was exposed in the residual boundary layer at night due to its high elevation, constantly replenished with the regional pool of air from daytime.
Overall, chemical transformation contributed ∼ 60 % of the daytime difference in GOM between AI and the two sites over land (TF and PM), 33 % of the nighttime difference between AI and TF, and 26 % of the difference between PM and AI.
Dry deposition and gas-particle partitioning contributed 4-37 and 30-96 %, respectively of the total GOM difference between AI and PM.Both processes had larger contributions at night that during daytime.Dry deposition contributed 6-24 % of the GOM difference between AI and TF and gas-particle partitioning 18-78 %.

Br chemistry in the MBL
Diurnal cycles of Br and BrO radicals (Fig. 6) were simulated using the Br chemical mechanism described in Sect. 2. Photodissociation of Br 2 was the main source of Br and BrO radicals during daytime.Our simulations suggested that reactive Br compounds were significant gaseous oxidants of GEM in the MBL at a fixed initial mixing ratio of 5.6 ppqv for Br 2 .Increasing initial mixing ratios of Br 2 by 25 % resulted in an increase of 0.01-2.15ppqv in GOM mixing ratios.
In addition, the reaction of BrO with methyldioxy (CH 3 O 2 ) radicals could have important influence on the mixing ratios of Br, BrO, and GOM.Simulated daytime mixing ratios of CH 3 O 2 was ∼ 40 pptv, and the rate coefficient of (5.7 ± 0.6) × 10 −12 cm 3 molecule −1 s −1 at 298 K for BrO + CH 3 O 2 (Aranda et al., 1997a) was used for our simulations.Pathways B1, B2, and B3 were suggested by Aranda et al. (1997a) based on an experimental study (Table 4).However, the production of CH 3 O may be due to its self-reaction in B1.Guha and Francisco (2003) suggested CH 3 OOOBr 15 Introduction

Conclusions References
Tables Figures

Back Close
Full to be a likely intermediate of this reaction, and that CH 3 OOOBr could dissociate to CH 2 O + HOOBr (B4, Table 4).Based on thermodynamics calculations, CH 3 OBr and O 2 (B3, Table 4) were possible products.BrOO and HOBr were both included in the Br chemical cycle and can be transformed back to Br and BrO radicals in the model.However, it is unclear whether CH 3 OBr (product of B3) or HOOBr (product of B4) could be transformed back to Br and BrO radicals in the atmosphere.In this case, using the B3 or B4 pathway did not appear to make a difference in our box model results.
In this study, the B1 and B2 pathways were used for the CH 3 O 2 + BrO reaction as part of the base case simulation (denoted as Sim-avg BrOO).The sensitivity case Sim-avg CH 3 OBr used the B3 pathway in lieu of B1 and B2.The simulated average and the range of GOM diurnal cycles in the base case and the sensitivity case were evaluated against observed mean and median GOM diurnal cycles of the 50 study cases at AI (Fig. 7).If the CH 3 O 2 + BrO reaction followed the B1 and B2 pathways, this reaction had a negligible effect on reactive Br radicals.However, if B3 or B4 was applied, the simulated total GOM mixing ratio was lowered by 50 % during daytime.
Moreover, the simulated GOM diurnal cycle in the base case agreed favorably with the observed average GOM diurnal cycle (NMSE = 15 %), while the results of the Sim-avg CH 3 OBr case were in better agreement with the observed median GOM diurnal cycle (NMSE = 14 %).These agreements indicated that, if the BrO + CH 3 O 2 reaction was a net sink of BrO radicals, the model was able to simulate better most cases, whereas if the product of BrO + CH 3 O 2 was transformed back to Br or BrO radicals, the model appeared to capture those cases with large GOM mixing ratios (> 6 ppqv).Due to the scarcity of kinetic research on the B3 and B4 pathways, we used B1 and B2 pathways for CH 3 O 2 + BrO reaction in this study.
In summary, the pathways of BrO + CH

Model evaluation
For all cases at AI and TF, the average simulated and observed GOM diurnal cycles agreed reasonably well in both magnitude and shape, whereas at PM the model appeared to have missed both (Fig. 8).Three salient features were noted for the disagreement between the model and observational results.First, standard deviation from observations was a factor of 2-7 larger than the simulated.This suggested that the model could capture the mean values of GOM, but not the very low and very large mixing ratios.Second, observed nighttime GOM mixing ratios were 12-200 % larger than the simulated at AI, indicating that the model did not capture certain nighttime processes producing GOM in the marine boundary layer.Third, the simulated diurnal cycle was the opposite of the observed at PM, with the maximum during the day and minimum at night.It was likely that the model simply simulated the dependence of GOM production on solar radiation.At PM, more processes contributed to the diurnal variation.At night, the site is above the nocturnal boundary layer and exposed to the GOM produced from the daytime, which could continually replenish surface GOM at the site that was lost via dry deposition and perhaps reduction.The model-observation discrepancies for the three sites were discussed as follows.

Appledore Island (marine)
Of the 50 cases at AI, 27 diurnal cycles of GOM were simulated with the average values and patterns close to the observed and NMSE s = 2.86 %, denoted as matching cases hereafter, 8 were underestimated with NMSE s = 146 %, and 15 were overestimated with NMSE s = 167 %.The observed and simulated average GOM mixing ratios and the corresponding ranges were calculated for the matching, under-estimation, and over-estimation cases at AI (Fig. 9a).For more than half of the time (27 matching cases out of 50 cases in total), the model captured the average GOM diurnal cycle, the diurnal cycle pattern and overall GOM levels.Beyond that, Fig. 9a shows large difference in the observed GOM levels among the matching, under-estimation, and over-estimation Introduction

Conclusions References
Tables Figures

Back Close
Full cases.On average, the observed daytime peak in the under-estimation cases was about twice as large as that for the matching cases and 7 times larger than that for the over-estimation cases.However, such difference was not captured by the model, suggesting that one or more GOM producing processes in the MBL were not included or not realistically represented in the box model.In addition, the GOM diurnal pattern of the over-estimation cases was different from those in the under-estimation and matching cases.The average observed GOM diurnal cycles of the under-estimation and matching cases both exhibited a daily maximum at 13:00 EDT and a minimum over 04:00-08:00 EDT whereas for the over-estimation cases a daily maximum at around 20:00 EDT and a minimum at 07:00-08:00 EDT.
Such differences were caused possibly by the challenges of simulating Br and BrO in the MBL at AI.No measurements of Br and BrO radicals as well as Br 2 were available at AI.To reasonably simulate mixing ratios of Br and BrO, Br 2 mixing ratios were calculated based on the BrO observations at a mid-latitude MBL site from Saiz-Lopez et al. (2006), which was ∼ 5.6 ppqv during the daytime (06:00-21:00 EDT).Saiz-Lopez et al. (2006) showed that the daytime peak mixing ratios of BrO in the MBL could vary by a factor of 2 over a time period of 3 days.Such variation was not captured in our box model, which could result in uncertainty of up to 100 % in simulated Br mixing ratios with subsequent effects on GOM simulation.
In the over-estimation cases, the simulated GOM daytime peaks were very low, and appeared later during the day than in the under-estimation and matching cases.Considering the late afternoon peak (17:00 EDT) of O 3 compared to the noontime peak of Br radicals, O 3 possibly played a more important role in the over-estimation cases.
To verify this hypothesis, a sensitivity simulation was conducted without the initial Br 2 mixing ratio fixed for these cases, termed as the O 3 / OH case.In this case, the Br 2 concentration rapidly diminished with time and consequently the concentrations of Br and BrO were very low.The O 3 / OH case turned out to better represent these 15 overestimation cases with NMSE s = 34 % (compared to 167 % with Br 2 mixing ratio fixed).Introduction

Conclusions References
Tables Figures

Back Close
Full This suggested that in the MBL, Br may be a dominant GEM oxidation most of the time, but occasionally at low Br mixing ratios, O 3 could become dominant.
To identify the potential sources of GOM at AI, backward trajectory analysis was conducted using the HYSPLIT4 model (https://ready.arl.noaa.gov/HYSPLIT.php).All 24 h backward trajectories started from the time of GOM daily peaks for the 50 cases.The trajectory results were clustered for over-estimation, matching, and under-estimation cases (Fig. 10).Based on these trajectories, in about half of the 15 over-estimation cases air masses originated from marine environments, while in more than 80 % of the 27 matching cases and 7 out of 8 under-estimation cases air masses came from inland northwest of AI.Note that in those under-estimation cases GOM mixing ratios were exceptionally large, exceeding 30 ppqv.
Different source areas of air masses reaching AI could be one of the reasons for the large variation of GOM observations.The highest levels of GOM were observed in summer with RH roughly < 50 % at AI (Mao et al., 2012).A close examination of the 50 cases at AI revealed low RH levels (≤ 45 %) on 16 days.The time periods with RH ≤ 45 % appeared mostly (78 % of the time) in the afternoon over 12:00-20:00 EDT and less so (22 %) at night over 21:00-02:00 EDT.During these time periods, increased GOM (15 out of 16, compared with periods with high RH on the same day) and daily maximum GOM (10 out of 16) occurred simultaneously at low RH, regardless of the time of the day.
Interestingly, the RH level of 45 % corresponds to the crystallization point of NaCl (Cziczo et al., 1997;Tang et al., 1997).The crystallization of sea-salt aerosols might be link to the very high GOM peaks in certain ways.Rutter and Schauer (2007) found that particles of potassium and sodium chlorides had high partitioning coefficients that could shift the GOM gas-particle partitioning toward the aqueous phase, while ammonium sulfate, levoglucosan, and adipic acid would shift the partitioning toward the gas phase.It was thus hypothesized that certain processes might have been activated during transport of inland air masses to the MBL involving the interaction between land and marine air, which potentially resulted in those very high GOM mixing ratios.Introduction

Conclusions References
Tables Figures

Back Close
Full  2012) found effective reactivity of chloride (Cl − ) components with organic acid in sea salt aerosols (SSAs), possibly leading to depletion of Cl − and formation of organic salts in aerosols.Biogenic compounds in air masses originating from inland forested areas could be oxidized forming organic acids in transit.As inland air reached the MBL, these organic acids would deposit onto SSAs and could subsequently change SSAs' chemical and physical properties, such as lowering concentrations of Cl − and forming a thick organic film on the outside of SSAs.The lower concentrations of Cl − and higher concentrations of organic acid in aerosols might have contributed to the shift in the gas-to-particle partitioning to the gas phase and resulted in higher GOM mixing ratios in the atmosphere.
Another possible explanation could be air masses of inland origin encountering marine air rich in atmospheric Br and BrO radicals.The main source of atmospheric Br is thought to come from the release of Br 2 and BrCl from SSA (Finlayson-Pitts, 2010;Sander et al., 2003).Experimental studies suggested Br − enhancements of a factor of 40 to 140 on the surface of sufficiently dry artificial SSA (Ghosal et al., 2008;Hess et al., 2007).Therefore, when drier inland air masses were mixed with marine air in the MBL under relatively low RH conditions, SSA became drier, forcing more Br 2 to be released from SSA, resulting in enhanced oxidation of GEM by Br and BrO radicals.These hypotheses need to be validated in future research.These mechanisms are presently missing in the box model, leading to the model's inability to capture very high GOM mixing ratios.Measurements of halogen species and a better gas-particle partitioning mechanism are needed to better the model's performance.

Thompson Farm (coastal)
Generally, the box model performed well at TF (Fig. 8b) with overall NMSE s = 0.75 % and RMSE = 0.78 ppqv.Of the 12 cases at TF, 7 diurnal cycles of GOM (58 %) were simulated reasonably well with NMSE s < 50 %, only one was underestimated by ∼ 70 %, and 4 cases were overestimated by a factor of 3 to 6. Overall, the observed average diurnal cycles of GOM for all selected summer clear-sky days at TF 20 Introduction

Conclusions References
Tables Figures
For the over-estimation and matching cases, the model reproduced very low GOM mixing ratios at night (Fig. 9b).For the same reason substantially lowering GEM mixing ratios at night and in the early morning at TF (Mao et al., 2008) Another notable feature in Fig. 9b is the exceedingly high observed GOM mixing ratios in the sole under-estimation case and the low observed GOM mixing ratios throughout the day in all over-estimation cases.Observed GOM mixing ratios in the under-estimation case showed a factor of 3-7 larger than those in the matching cases, and a factor of 3-31 larger than those in the over-estimation cases (Fig. 9b).Concurrently, larger fine particle concentrations, 8272 cm −3 on average, were observed for the under-estimation case, which was 65 and 93 % larger than those in the matching cases and over-estimation cases, respectively.Lower RH, 59 % on average, was observed in the under-estimation case, 11 and 15 % lower than that in the matching and over-estimation cases, respectively.Moreover, higher air pressure (1017, 7 and 10 hPa larger than the matching and over-estimation cases, respectively), lower wind speed (0.59 m s −1 on average, 47 and 56 % lower than matching and over-estimation cases respectively), and stronger solar radiation flux (8 and 13 % stronger than matching and over-estimation cases respectively) were found in the under-estimation case.Clearly, the under-estimation case occurred under the strongest Bermuda High influence, with Introduction

Conclusions References
Tables Figures

Back Close
Full the calmest, sunniest, and driest conditions of all cases, which is most conducive to photochemistry and pollution build-up that may have ultimately contributed to the very large GOM mixing ratios in that one under-estimation case.Our model appeared to fail to mimic the chemistry under such conditions that produced the largest GOM mixing ratios.

Pack Monadnock (inland, rural, elevated site)
At PM, diurnal cycles of GOM were overestimated with NMSE s = 70 % and overall RMSE = 0.13 ppqv.However, considering the extremely low mixing ratios of GOM observed at PM (Fig. 2), cases with RMSE < 0.1 ppqv (LOD) were considered as matching cases.Therefore, the model reasonably simulated 11 out of 21 (52 %) cases, underestimated in 3, and overestimated in 7. Evaluation of simulated GOM diurnal cycles against observations (Fig. 8c) showed reasonable agreement with general overestimation ranging over 0.05-0.07ppqv.
In comparison, rest (86 %) of the cases were showing a very flat pattern of GOM diurnal cycle at PM.The first and the most important reason for such observation-model discrepancy is that the PM site is a mountain site (700 m a.s.l.), which is above the nocturnal inversion layer (∼ 200 m at TF) at night but in the middle of the convective boundary layer during the day.At night, a regional pool of GOM produced during daytime remained in the residual layer, which kept the surface GOM levels from dropping below the LOD at night at PM.The slight decline of GOM mixing ratios after sunrise was because of mixing with the lower altitude air masses with depleted GOM from the night.The effect of the PM's site characteristics was not represented in the box model, which could result in model's inability to simulate diurnal variation associated with this aspect of the site.In addition, due to the dominance of GEM oxidation by O 3 in GOM production in the model, it was highly likely that the flat patterns (slightly higher at night) of GEM (Fig. 2) and O 3 were mirrored in GOM mixing ratios.

Conclusions References
Tables Figures

Back Close
Full

Summary
This study provided a state-of-the-art chemical mechanism for atmospheric Hg modeling system and tested the chemical mechanism for three different environments using a mercury box model.Eighty-three summer clear-sky days were selected at marine, coastal, and inland elevated sites in southern New Hampshire to evaluate the model.
As a result, for each of the three environments, GOM diurnal cycles of over half selected cases were reasonably represented by the box model.It was hypothesized that dry air masses with organic compounds transported from inland may result in very large GOM mixing ratios in the MBL possibly due to changing physical and chemical properties of sea salt aerosols.The low nighttime and morning GOM mixing ratios at coastal site were likely a result of a net loss due to dry deposition in the nocturnal inversion layer.The GOM mixing ratios above the limit of detection at the inland site at night were probably caused by constant replenishment from a regional pool, in the residual boundary layer, of GOM that was produced in the daytime convective boundary layer.The updated chemical mechanism largely improved GOM diurnal cycle simulations at the coastal and inland sites.HgO produced from oxidation of GEM by O 3 and OH dominated GOM species at the coastal and inland sites, while bromine-induced mercury species (mainly BrHgOOH, BrHgOBr, and HgBrO) were important at the marine site.In Br chemistry, the products of the CH 3 O 2 + BrO reaction strongly influenced the simulated Br and Hg concentrations.In this study, GEM oxidation by O 3 and OH was represented in ways similar to those in regional and global models, which is limited by the current nebulous understanding of potential surface chemistry.It should also be acknowledged that studies have suggested problems in GOM measurements using the current Tekran instruments (Gustin et al., 2015;Jaffe et al., 2014).If indeed real atmospheric GOM concentrations were underestimated in Tekran measurements, it implies even more confounding, yet unknown issues in our current understanding of Hg chemistry.More experimental or theoretical studies on Hg reactions and better GOM measurement data are warranted to improve our understanding and subsequently model Introduction

Conclusions References
Tables Figures

Back Close
Full    Full . Deposition of atmospheric Hg is one of the most important sources of aquatic Hg. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

1.
Oxidations of Hg by O 3 , OH, HOCl, and ClO − , further oxidation of HgOH by O 2 ; 2. Reduction of Hg 2+ by HO 2 , photolytic reduction of Hg(OH) 2 and S(IV)-mediated reduction; and 3. Aqueous equilibria involving HgSO 3 , Hg(SO 3 ) + and Hg(OH) 2 .Gasphase halogen reactions in the box model are mainly cycles of halogen radicals (Cl / Br / I and ClO / BrO / IO radicals).The Cl / Br / I radical cycles include photodissociation of Cl 2 / Br 2 /I 2 , organic halides, and other inorganic halides as sources, and oxidation reactions as sinks.The ClO / BrO / IO radical cycles involve oxidation of Cl / Br / I radicals, photodissociation of ClNO 2 / ClONO 2 / BrNO 2 / BrONO 2 , production from other halogen radicals, and sink reactions to reproduce Cl/Br/O radicals or other halides.Aqueous halogen reactions include reactions of Br − / Cl − and reactions of aqueous BrCl, HCl, HBr, HOCl, HOBr, Cl 2 , and Br 2 species.The chemistry of halogen radicals, especially the reaction cycles of Br and BrO radicals, could be important and should not be neglected Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | National Oceanic and Atmospheric Administration (NOAA) Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) trajectory model was used to identify source regions of air masses at the three sites.The model runs were performed over twenty-four hours using the NOAA NAM (Eta) Data Assimilation System (EDAS) data with a 40 km×40 km horizontal resolution as input.Backward trajectories and trajectory clusters were calculated.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Cases 5-8.Using the slowest rate coefficient of GEM + O 3 obtained fromHall (1995), as opposed to the one fromSnider et al. (2008) led to a decrease of 56.7 % in HgO, and decreases of 15 and 85 % in total GOM during daytime and nighttime, respectively.Turning off GEM oxidation by O 3 , OH, or Br resulted in decreases of 19, 10, and 30 %, respectively, in daytime GOM mixing ratios.Turning off the GEM + Br oxidation reaction also decreased daytime PBM mixing ratios by 45 %.However, for nighttime GOM and PBM mixing ratios, turning off the GEM + O 3 reaction caused decreases of 92 and Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 3 O 2 could play an important role in atmospheric Br chemistry and Hg speciation in Br-rich environments.Research on the reaction pathways and rate coefficients of the BrO + CH 3 O 2 reaction is warranted to better assess the role of this reaction.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Laskin et al. ( Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | simulations of atmospheric Hg cycling, which can ultimately serve policy-making in an effective mannerDiscussion Paper | Discussion Paper | Discussion Paper | Photochemistry 1 No Yes 3.0 × 10 −11
Sigler et al. (2009)ndSigler et al. (2009).Table2lists the input variables of the box model.The model's initial mixing ratios of GEM, O 3 , CO, and NO were obtained from observations and were set to be constant during a simulation.Dry deposition flux was calculated using dry deposition velocity data derived from Zhang , the low nighttime GOM at TF was probably caused by loss via dry deposition under nocturnal inversion.To capture these low values in model simulations, realistic nocturnal boundary layer height data were needed beside solid representation of dry deposition and chemistry in the model.The diurnal cycle of boundary layer height in the box model was parameterized based on reanalysis data obtained from the Research Data Archive at the National Center for Atmospheric Research (http://rda.ucar.edu/datasets/ds093.0/).Use of these data helped to reproduce the low nighttime GOM levels in simulations for the TF site.

Table 1 .
Gas phase Hg reactions in the box model.