Article pubs.acs.org/est
Parameterizing Soil Emission and Atmospheric Oxidation-Reduction in a Model of the Global Biogeochemical Cycle of Mercury Tetsuro Kikuchi,*,† Hisatoshi Ikemoto, Katsuyuki Takahashi, Hisashi Hasome, and Hiromasa Ueda Japan Environmental Sanitation Center, 10-6, Yotsuya-kamicho, Kawasaki-ku, Kawasaki, Kanagawa 210-0828, Japan S Supporting Information *
ABSTRACT: Using the GEOS-Chem atmosphere−land−ocean coupled mercury model, we studied the significances of two processes, soil emission and atmospheric oxidation−reduction, in the global biogeochemical cycling of mercury and their parametrization to improve model performance. Implementing an empirical equation for soil emission flux (Esoil) including soil mercury concentration, solar radiation, and surface air temperature as parameters enabled the model to reproduce the observed seasonal variations of Esoil, whereas the default setting, which uses only the former two parameters, failed. The modified setting of Esoil also increased the modelsimulated atmospheric concentration in the summertime surface layer of the lower- and midlatitudes and improved the model reproducibility for the observations in Japan and U.S. in the same period. Implementing oxidation of atmospheric gaseous elemental mercury (Hg0) by ozone with an updated rate constant, as well as the oxidation by bromine atoms (Br) in the default setting, improved the model reproducibility for the dry deposition fluxes observed in Japan. This setting, however, failed to reproduce the observed seasonal variations of atmospheric concentrations in the Arctic sites due to the imbalance between oxidation and reduction, whereas the model with Br as the sole Hg0 oxidant in the polar atmosphere could capture the variations. water than Hg0.4 As a consequence, Hg(II) and Hgp are quickly removed from the atmosphere by wet and dry deposition processes.2,4 A portion of the Hg that is deposited into water bodies can be transformed to neurotoxic methyl Hg.7 A number of numerical models for simulating the abovementioned processes of Hg on regional and/or global scales have been developed and used to assess the impact of anthropogenic and natural Hg emissions on the environment and human health. A global atmosphere−land−ocean coupled Hg model based on the global atmospheric chemical transport model (CTM) GEOS-Chem8 (hereafter called “GEOS-ChemHg”) has been developed and utilized to estimate the biogeochemical cycling of Hg and assess the fate of anthropogenically emitted Hg in the environment on the global scale.1,9−12 We have used GEOS-Chem-Hg to study the significance of each process of Hg in the environment, as well as the optimization of its parametrization. In this study, we focused on two processes: emission from soil and oxidation− reduction in the atmosphere. Emissions of Hg0 from the surfaces and its oxidation rate in the atmosphere are considered to be the fundamental parameters that determine the air− surface exchange of Hg, especially in areas remote from the major anthropogenic sources.6
1. INTRODUCTION Historical anthropogenic emissions of mercury (Hg) are considered to have increased the amount of Hg cycling on the Earth’s surface and hence increased the health risk this element poses to living organisms.1,2 The fate of Hg in the environment is determined by several physical and chemical processes, including emission, transport, chemistry, and deposition. Mercury is emitted from volcanic activities and volatilized from land (soil and vegetation) and water (ocean, lakes and rivers) surfaces to the atmosphere as gaseous elemental Hg (Hg0), although significant quantities of divalent Hg (Hg(II)) are also likely formed at volcanic vents as gaseous HgCl2 when magmatic gases are cooled and oxidized.3 Anthropogenic sources emit not only Hg0 but also oxidized Hg species (mostly Hg(II)) in the gas phase and Hg(II) associated with particulate matter (Hgp).2,4 Currently, estimates of anthropogenic emissions of Hg are considered to be more reliable than those of natural emissions of Hg0 from the surfaces, whereas the latter emissions are roughly half to twothirds of total global Hg0 emissions.5,6 In the atmosphere, Hg0 is slowly oxidized to Hg(II) and/or Hgp, and Hg(II) reduction to Hg0 can also occur in cloudwater droplets. Elemental Hg has a high vapor pressure and a low water solubility, which together with a slow oxidation rate, favors its long-range atmospheric transport from an emission source, as well as revolatilization of previously deposited Hg as Hg0.4 On the other hand, Hg(II) species such as HgCl2 generally have lower vapor pressures and are more soluble in © 2013 American Chemical Society
Received: Revised: Accepted: Published: 12266
March 12, 2013 September 20, 2013 September 23, 2013 September 23, 2013 dx.doi.org/10.1021/es401105h | Environ. Sci. Technol. 2013, 47, 12266−12274
Environmental Science & Technology
Article
Holmes et al.9 (especially for the atmospheric CTM part) and by Soerensen et al.10 (especially for the ocean model part). The 2005 global inventory prepared by AMAP and UNEP4 with a 1° × 1° horizontal resolution, which has been already arranged as an input data for GEOS-Chem-Hg, was used as the value for anthropogenic Hg emissions to the atmosphere, although the Hg0: Hg(II): Hgp speciation for fossil fuel emissions, which make up 46% of the global total anthropogenic Hg emissions, was changed from 50:40:10 to 86.5:9.9:3.6.22 In addition to anthropogenic emissions, GEOS-Chem-Hg includes the following Hg emission processes: prompt reemission of wet and dry deposited Hg(II) and Hgp, emission from geogenic sources (not including volcanic emissions), emission due to biomass burning, soil emission, and emission from melting snow and ice. The Hg species emitted through these processes is assumed to be entirely Hg0. Mercury emission from geogenic sources, amounting to 500 Mg yr−1 as a global total, is spatially distributed according to the location of Hg mines as an indicator of Hg deposits,1 although the default setting reduces this flux by 50%.10 In this study, we returned this flux to the original estimate, which is consistent with the estimated emission flux from desert/metalliferous/nonvegetated zones by Pirrone et al.5 (546 Mg yr−1). Mercury emission flux due to biomass burning is calculated by multiplying an inventory of carbon monoxide (CO) emission resulting from biomass burning by the Hg/CO molar ratio in the biomass burning plume. The Global Fire Emissions Database (GFED) can be used for the CO emission in GEOS-Chem-Hg, and we chose the monthly data from GFED version 3.23 For the Hg/ CO ratio, we used the world average estimated by Friedli et al.24 (154 nmol mol−1) instead of the default setting (100 nmol mol−1).9 The Hg/CO ratio by Friedli et al.24 produced 309 MgHg yr−1 as the global total flux in 2008, whereas the default setting calculated the flux as 201 MgHg yr−1. In GEOS-ChemHg, emitted Hg from anthropogenic and natural sources is assumed to be distributed throughout the boundary layer. GEOS-Chem-Hg handles wet scavenging of Hg(II) and Hgp following the scheme of Liu et al.25 and dry deposition of Hg0, Hg(II) and Hgp following the resistance-in-series scheme of Wesely.26 Uptake of Hg(II) by sea-salt aerosol and its subsequent deposition in the marine boundary layer (MBL) is also calculated separately.9 Gaseous Hg(II) is assumed to be wet scavenged as HgCl2. As the Henry’s law constant of HgCl2 (KH_HgCl2), which determines gas−liquid water partitioning of Hg(II), we used the Arrhenius expression proposed by Sommar et al.27 (eq 1 below) instead of the default constant value (7.1 × 10−7 atm L mol−1 obtained at 298.15 K).28,29
The understanding of which and how environmental factors control the soil Hg emission is important for properly managing Hg.13,14 The current version of GEOS-Chem-Hg calculates the soil Hg emission flux as a function of soil Hg concentration and solar radiation.9 On the other hand, it has been reported that this flux can be expressed as an Arrhenius equation (i.e., temperature dependent).15−17 Furthermore, Lin et al.18 found a synergistic effect between light and air temperature on the soil−Hg emission flux, and they attributed this synergy to the kinetic enhancement of Hg(II) photoreduction in soil at higher temperature. We implemented the empirical equation proposed by Lin et al.,18 which uses not only soil Hg concentration and solar radiation but also air temperature as parameters, and compared the results to the computations of the default setting. Many of the existing Hg atmospheric CTM assume that ozone (O3) and the OH radical are the primary oxidants of Hg0 in the atmosphere, although there have been debates about the accuracy of their proposed reaction rate constants and the significance of these reactions in real atmospheric condition.19 On the other hand, Hg0 oxidation by bromine atoms (Br) is reported in numerous studies as the likely primary cause of the depletion of Hg0 in the Arctic boundary layer in spring known as the atmospheric mercury depletion events (AMDEs).20 Holmes et al.9 found that GEOS-Chem-Hg could reproduce the springtime depletion and summer rebound of total gaseous Hg (the sum of gaseous Hg0 and Hg(II)) observed at the polar sites when Br is assumed to be the sole oxidant of Hg0. Such a setting of GEOS-Chem-Hg, however, could not capture the summer maximum of the Hg wet deposition flux over the southeastern U.S. because of the low Br concentration in the subtropics. As the default setting, the current GEOS-Chem-Hg assumes Hg0 oxidation by Br alone, and the oxidation reactions by OH, O3 and bromine oxide (BrO) can also be implemented as options in the model. We include O3 as an oxidant of Hg0 in addition to Br, and assess how the new chemistry influences the model performance.
2. MATERIALS AND METHODS 2.1. Model Simulation Condition and Base Model Setting. In this study, we used GEOS-Chem-Hg version 9-0102.21 The GEOS-Chem model is driven by assimilated meteorological data from the Goddard Earth Observation System (GEOS) of the NASA Global Modeling and Assimilation Office. We used the GEOS-5 meteorological fields, which originally have a horizontal resolution of 1/2° latitude by 2/3° longitude, with 72 hybrid eta levels from the surface to 0.01 hPa. For computational efficiency, the meteorological input data to the model was degraded to 2° × 2.5° horizontally and 47 levels vertically. We used the atmosphere and the surface ocean conditions at the beginning of 2008 as the initial conditions of the Hg fields, which had been already calculated from a GEOS-Chem-Hg simulation with the default setting and stored in the prescribed files. A set of model simulations was driven by the meteorological field in 2006−2009, and the output for 2007−2009 was used for analysis. (The initial year (2006) was regarded as the model spin-up period.) The gchem version 0.1.1, a package of Python routines (https://github.com/gkuhl/gchem/), was used to read the model output files and create the input files. GEOS-Chem-Hg addresses three types of Hg species: Hg0, Hg(II) and Hgp. The calculation conditions and the scientific bases of the default GEOS-Chem-Hg are described in detail by
KH HgCl2[atm L mol−1] = 5.53 × 105 × exp( −67.2 × 103/RT ) −1
(1) −1
where R is the gas constant (8.314 J K mol ), and T is air temperature (K). For the Henry’s constant of Hg0 used to calculate its dry deposition flux, we also updated the default value (0.11 mol L−1 atm−1 obtained at 298.15 K)29 to the value derived from the Arrhenius expression for pure water by Andersson et al.30 (0.319 mol L−1 atm−1 obtained at 298.15K). The 50/50 partitioning of atmospheric Hg(II) between the gas and aerosol phases was assumed by default when the dry deposition rate was calculated.9 (The dry deposition rate was calculated as an average of the gaseous and aerosol deposition rates.) Numerous observations, however, have shown that ambient concentrations of Hgp are generally higher than those 12267
dx.doi.org/10.1021/es401105h | Environ. Sci. Technol. 2013, 47, 12266−12274
Environmental Science & Technology
Article
of gaseous Hg(II) (e.g., Weigelt et al.31). Amos et al.32 have derived from the observational data an empirical gas-particle partitioning relationship of Hg(II) including fine particulate matter (PM2.5) concentration and temperature as parameters, which was not validated in the present study but has been implemented by default in the newest version of GEOS-ChemHg (v9-01-03).21 Hereafter we refer to the model with the default settings and the above-mentioned modifications to the geogenic and biomass burning emission fluxes and the Henry’s constants of atmospheric Hg0 and Hg(II) as “BASE”. 2.2. Parameterization of Soil Hg Emission Flux. By default, GEOS-Chem-Hg calculates the soil Hg emission flux (Esoil) as a function of soil Hg concentration and solar radiation as follows:9 Esoil[ng m−2h−1] = fsoil Csoil exp(1.1 × 10−3 × LAL )
those over the Japanese Archipelago and the Korean Peninsula. In this study, we replaced the zero Csoil for each land-grid over Japan (I−VII in Figure S1 in the Supporting Information (SI)) by the median of the total Hg concentrations in the river sediment samples collected at the locations in each grid (n = 163−356 per grid).35 Although a river sediment sample consists of not only the surface soil but also the bedrock and sediment distributing in the upstream watershed, the range of the Hg concentrations in the Japanese river sediments referenced in this study (54.0 ± 46.8 ng g−1 as the mean ± a standard deviation; n = 3024)35 can be assumed to be comparable to that of the surface soils (290 ± 460 ng g−1; n = 469)36 (see the discussion in SI Section S2.1). We also adopted the average of the total Hg concentrations in river sediments of Korea (55.5 ng g−1, n = 46)37 as the Csoil for the grids over the Korean Peninsula (K-1 and K-2 in SI Figure S1) instead of zero. 2.3. Parameterization of Oxidation−Reduction of Atmospheric Hg. The default GEOS-Chem-Hg assumes Br to be the sole oxidant of Hg0 in the atmosphere.9 In M2-1 and M2-2, we added the oxidation by O3 as well. For the oxidation rate constant by O3 (kox_O3), the following Arrhenius expression by Pal and Ariya38 was implemented instead of a constant value (3.0 × 10−20 cm3 molecule−1 s−1 obtained at 293.15 K)39 as the optional setting of GEOS-Chem-Hg:
(2)
where fsoil is the scaling factor of soil Hg emission (=5.512 × 10−2 g m−2 h−1), Csoil is soil Hg concentration (ng g−1), L is shortwave solar radiation (W m−2), and AL is the fraction of light attenuated by leaf canopy before reaching to the ground, which is parametrized by leaf area index (LAI).1 Instead of this parametrization, in the other three models in this study (“M1”, “M2−1” and “M2−2”), we implemented the empirical equation proposed by Lin et al.18 for Esoil from dry soils (∼0 wt% of soil moisture) with background Hg contents: −2 −1
−3
Esoil[ng m h ] = (10
kox O3[cm 3molecule−1s−1] = 8.43 × 10−17 × exp{(− 11700 ± 270)/RT }
× Csoil) × {β0 + β1Ts + β2LAL
2 2 + β3TLA L + β4 Ts + β5(LAL ) } s
In our simulations, the lower limit of the range of this rate constant, that is, 8.43 × 10−17 × exp(−11970/RT), was actually used. Hereafter we refer to the models with Br as the sole oxidant of Hg0, that is, BASE and M1, as the “Hg + Br models” and those with both Br and O3 as Hg0 oxidants, that is, M2-1 and M2-2, as the “Hg + Br/O3 models”. We did not include Hg0 oxidation by the OH radical because any proposed rate constants would be highly uncertain.19 The current GEOS-Chem-Hg includes the photoreduction of Hg(II) in liquid cloud droplets. The model assumes that the photoreduction rate constant (kred) is proportional to the photolysis frequency of nitrogen dioxide (JNO2 (s−1)) and is calculated as follows:9
(3)
where Ts is air temperature at 2 m above the ground (K), β0 = 4.12 × 102, β1 = −53.9, β2 = 4.26 × 10−1, β3 = −1.78 × 10−3, β4 = 1.81, and β5 = −4.85 × 10−4. Esoil in eq 3 can be regarded as a quadratic function of Ts, and it reaches the minimum (= 0) when Ts is 287.15 K. In M1, M2-1, and M2-2, the value of Esoil was fixed to 0 if Ts < 287.15 K. This setting agrees with the finding by Marumoto and Sakata32 from their laboratory experiments that there was no apparent Hg emission from soil when soil surface temperature was below approximately 288.15 K. Esoil in eq 3 can also be seen as a quadratic function of LAL, and it reaches the maximum when LAL is 400 W m−2. In calculation of Esoil in M1, M2-1, and M2-2, we fixed the value of LAL in eq 3 to 400 W m−2 if LAL > 400 W m−2. Soil moisture has been also reported to significantly enhance Esoil in field and laboratory studies.18,33,34 Lin et al.18 have proposed another empirical equation of Esoil using not only soil Hg concentration, solar radiation and air temperature but also soil moisture as parameters, whose implementation to the model has not been fully considered in this study. Csoil in each model grid is given as follows: Csoil = 45fsoildist
(5)
k red[s−1] = fred fa JNO2
(6)
where f red is the scaling factor of kred, fa is the fraction of Hg(II) which is partitioned to the aqueous phase. The value of f red used in each model was adjusted to match the simulated atmospheric Hg concentration to the observations (Table 1). An evaluation of the validity of f red in M2-1 and M2-2 (for the region between 60°S and 60°N) is stated in SI Section S3. In the polar regions, Br plays a significant role in determining the seasonal variation in atmospheric Hg concentration and deposition when compared to the role of O3.9,11,20 M2-2 omitted the Hg0 oxidation by O3 in the polar regions (south of 60°S and north of 60°N). In addition, the f red in these regions accordingly reduced to the same value as M1. The settings for Esoil and the oxidation−reduction of atmospheric Hg in each model are summarized in Table 1. 2.4. Validation of Model Performance. The performance of each model was verified by comparing the simulated Esoil, atmospheric Hg concentrations and its dry and wet deposition fluxes with observations. We evaluated the model reproducibility mainly for the seasonal and latitudinal variations of these observations.
(4)
where fsoildist is the spatial distribution factor of soil Hg concentration, which we used in all the models has been already calculated following the method of Selin et al.1 from a set of GEOS-Chem-Hg simulations with the default setting and saved in a prescribed file. (In this study, we gave priority to comparison between the performances of the default and modified parametrizations of Esoil using the same fsoildist values, although fsoildist should ideally be updated to match the parameter settings in each model.) However, we found that in the file zeros were allocated to some land-grids, including 12268
dx.doi.org/10.1021/es401105h | Environ. Sci. Technol. 2013, 47, 12266−12274
Environmental Science & Technology
Article
Table 1. Parameterizations of Soil Hg Emission flux (Esoil) and Oxidation-Reduction of Atmospheric Hg Implemented in Each Model
elevation of the site, whereas the simulated Hg0 in the MBL is for the surface layer. The simulated dry and wet deposition fluxes of Hg(II) and Hgp were validated based on the monitoring results in Japan and Europe.42−44 The dry deposition sampling at Japanese sites was carried out using a circular water surface sampler, which enables collection of only dry-deposited Hg(II) and Hgp (Hg0 is not collected).45
oxidation−reduction of atmospheric Hg model name BASE M1 M2-1 M2-2
model type by Hg0 oxidation Hg Hg Hg Hg
+ + + +
Br Br Br/O3 Br/O3
Esoil eq eq eq eq
2 3b 3b 3b
Hg0 oxidants
f reda
Br Br Br and O3 Br and O3 (except for polar regionsc) Br (polar regionsc)
3.5 × 10−4 1.75 × 10−4 1 × 10−1 1 × 10−1 (except for polar regionsc) 1.75 × 10−4 (polar regionsc)
3. RESULTS AND DISCUSSION 3.1. Soil Hg Emission Flux. Figure 1 (a) shows the temporal variations of the model-simulated monthly mean Esoil (bars) and Ts (brown solid line) averaged for the seven landgrids (I−VII in SI Figure S1) over Japan in 2007. The emission flux with the modified setting (eq 3; green bars) showed a clear seasonal variation with almost zero emission in winter and the maximum in summer, whereas the seasonal variation with the default setting (eq 2; blue dotted bars) was small. The emission flux with the modified setting was also strongly correlated with Ts (p < 0.001). This result is in accordance with the findings from the previous laboratory and field studies that Esoil is dependent on temperature.15−17,33 Figure 1 (b) illustrates the latitudinal distributions of the simulated global mean Esoil for the land-grids in summer months (June−August in the Northern Hemisphere (NH) and December−February in the Southern Hemisphere (SH)) of 2007−2009. The emission flux with the modified setting (green line) showed a characteristic distribution pattern in the both hemispheres: the maximum between the tropic and 35°, and then a steep decline in the higher latitudes. This distribution pattern basically corresponds to that of Ts, although the maxima around 30° would be attributed to the synergistic effect between temperature and solar radiation (SI Figure S2).18 Compared to the modified setting, the latitudinal variation of Esoil with the default setting (blue line) for the same summer months was very small. In addition, the summertime Esoil with the modified setting was an average of 3.1 times larger than the default setting, and 7.6 times greater at the maximum. The model-simulated Esoil is compared with the observations in Table 2.17,33,46−48 Both the model with the default setting and those with the modified setting generally underestimated the observed fluxes. This is mainly due to the substantially lower soil Hg concentrations used in the model simulations compared to the observed values (Table 2). The Esoil measured in Oak Ridge17 and Tahquamenon River watershed (TRW)46
a
Scaling factor of Hg(II) photoreduction rate constant (eq 6). bIf Ts < 287.15 K, Esoil is fixed to zero (Ts: air temperature at 2 m above the ground). If LAL > 400 W m−2, LAL in the equation is fixed to 400 W m−2 (L: shortwave solar radiation; AL: fraction of light attenuated by leaf canopy). cPolar region: < −60°N and > 60°N.
The simulated monthly mean total Hg (THg ≡ Hg0 + Hg(II) + Hgp) concentrations in the surface layer (ca. 130 m in height) of the land-grids over Japan (II−VII in SI Figure S1; 33−41°N) in 2007 were compared with the THg data of the Monitoring Results of Hazardous Air Pollutants compiled by the Japanese Ministry of Environment.40 The mean value of the monthly surface concentrations measured at all the monitoring stations located in each model grid (vertically the height of the station is lower than the mean surface height plus the surface layer height of the grid) was used for comparison. The data for the stations categorized as under direct influences from nearby stationary or mobile emission sources were excluded. The sampling duration and frequency of the observed THg were a single day and once a month, respectively. The data whose concentrations fell above or below Cave ± 3σ (Cave: monthly mean concentration; σ: standard deviation) for the grid were also excluded (see SI Section S4.1). The observed monthly mean atmospheric Hg0 concentrations at two midlatitude sites in U.S. and two Arctic sites (SI Table S2) and the Hg0 concentration in the MBL measured during a global circumnavigation41 (SI Table S3) were also referred. (All of these data can be downloaded from https://github.com/noelleselin/ HgBenchmark.) The simulated Hg0 values for the USA and Arctic sites are for the layer of model grid corresponding to the
Figure 1. (a) The mean values of the model-simulated monthly Esoil (bars) and the monthly mean Ts (brown solid line) for the seven grids over the Japanese Archipelago (I−VII in SI Figure S1) in 2007. A red dashed line represents the Ts below which the simulated Esoil by the modified setting (green bars) is fixed to zero. Error bars represent ± (a standard deviation) (the range of Esoil < 0 not shown). (b) Latitudinal distributions of the simulated global mean Esoil for the land-grids in the summer months (June−August in the NH and December−February in the SH) of 2007−2009. 12269
dx.doi.org/10.1021/es401105h | Environ. Sci. Technol. 2013, 47, 12266−12274
Environmental Science & Technology
Article
Table 2. Comparison of the observed (Obs), model-simulated (Model) and calculated (Calc) soil Hg emission fluxes (Esoil)a Esoil (ng m−2 h−1) surface soil Hg conc. (ng g−1) site
obs
Maebashi (Japan)
Jinyun Mountain (China)
Oak Ridge (USA)
forest open field forest open field forest open field forest open field Watson Forest
model
270
30
(no data)
28.4
modelb month or season Jun. Aug. Nov. spring
summer fall winter 469d
21.8
Apr.−Aug.
obs 4.0 12−14
