Journal of Geophysical Research: Atmospheres

A global model of the coupled sulfur/oxidant chemistry in the troposphere: The sulfur cycle

Authors


Abstract

[1] A sulfur cycle chemistry scheme with dimethyl sulfide (DMS), SO2, sulfate, H2S, and methanesulfonic acid (MSA) is included in the OsloCTM2 model, and concentrations of sulfur are calculated interactively with the oxidant chemistry. This allows more consistent estimates of aqueous phase oxidation of SO2 to sulfate by O3, H2O2, and HO2NO2. The year 1996 is chosen as the standard, and a model run with 1996 meteorology and emissions is compared with 1996 observations. The results agree well with observations overall, although the model tends to overestimate SO2 and underestimate sulfate in Northern Hemisphere winter owing to an oxidation limitation. A global budget for 1996 quantifying the various processes is investigated. Our model results give a global lifetime (global burden) of 1 day (0.25 Tg(S)) and 3.8 days (0.5 3 Tg(S)) for SO2 and sulfate. Differences between the Southern Hemisphere, characterized by natural emissions and by loss of SO2 by O3 and H2O2 oxidation, and the Northern Hemisphere, characterized by anthropogenic emissions and by large loss by dry deposition, are revealed. Significant changes in sulfur emissions have occurred over the last decades with decrease in the Unites States and Europe and increase in Southeast Asia. U.S., European, and Chinese SO2 emissions have changed by −17.6%, −47.5%, and +93%, respectively. To study the impact of emission changes on the atmospheric composition, we have calculated distributions using the Global Emissions Inventory Activity (GEIA) 1985 inventory. The changes in sulfur emissions have significant changes on the sulfur concentrations and also some effect upon the oxidants. Increased emissions of NOx and hydrocarbons in China enhance O3, but increased sulfur inhibit the increase. The SO2 oxidation by OH, which can lead to formation of new sulfate particles, is given special attention. The model run using GEIA 1985 anthropogenic emission inventory is compared with other model studies.

1. Introduction

[2] The sulfur cycle is one area in atmospheric chemistry where human activity has a large impact. The anthropogenic contribution is about three times larger than the natural in terms of emissions. Several studies have been performed over the last decades to investigate and quantify the processes of importance in the sulfur cycle [Rodhe and Isaksen, 1980; Langner and Rohde, 1991; Chin et al., 1996; Feichter et al., 1996; Kasibhatla et al., 1997; Restad et al., 1998; Roelofs et al., 1998; Koch et al., 1999; Barth et al., 2000; Iversen and Seland, 2002; Liao et al., 2003]. The COSAM exercise [Barrie et al., 2001; Lohmann et al., 2001; Roelofs et al., 2001] provides an excellent overview of different sulfur models, and comparison both between models and between models and observations. Earlier studies on the sulfur cycle focused on the acidification of the atmosphere, but lately there has been more concern about the sulfate particles formation and their impact on climate. Sulfate particles affect the radiative balance of Earth in two ways: The direct radiative effect, where the sulfate particles scatter solar radiation, and the indirect effects where the sulfate particles: (1) reduce the cloud droplet size and change the optical properties of clouds and (2) reduce the precipitation efficiency and increase cloud water content and lifetime of clouds. Both the direct and indirect effects of sulfate aerosols contribute to a cooling of Earth's surface [Intergovernmental Panel on Climate Change (IPCC), 2001].

[3] The main constituents in the tropospheric sulfur cycle are DMS (dimethylsulfide, CH3SCH3), SO2 (sulfur dioxide) and sulfate (SO42−). In addition H2S (hydrogen sulfide) and MSA (methanesulfonic acid, CH3SO3H) contribute to the sulfur cycle. Emissions of OCS are small although it is the most abundant sulfur specie in the atmosphere due to its long lifetime in the troposphere [Kjellström, 1998]. Quantitatively the most important emissions in the sulfur cycle are the anthropogenic emissions of SO2 and the natural emissions of DMS. Natural emissions of DMS are connected with large uncertainties and need to be better quantified [Boucher et al., 2003]. The changes in anthropogenic emission patterns of SO2 show large variations where anthropogenic emissions over Europe and the United States have decreased over the last decades and emissions in Asia have increased. The significance of these changes needs to be studied.

[4] To study the sulfur cycle, both chemical tracer models (CTMs) and general circulation models (GCMs) are used. Simple sulfur cycle parameterizations have been widely applied using off-line calculations of chemical oxidants. However, some former studies apply semiprognostic oxidants, i.e., that prescribed 5-days averages of J values and concentrations of OH and HO2 are used to calculate concentrations of H2O2 [Koch et al., 1999; Barth et al., 2000], Roelofs et al. [1998] implemented a fully coupled scheme allowing chemical feedback and Liao et al. [2003] investigate interactions between ozone-NOx-hydrocarbons chemistry and aerosols. In this study with the OsloCTM2 model [Sundet, 1997] an interactive tropospheric sulfur and oxidant chemistry scheme is applied. Hence the sulfur species and the oxidants (OH, O3, H2O2, HO2NO2, NO3) are calculated simultaneously. This sulfur-oxidant interaction is an improvement compared to previous off-line model studies as it includes chemical interactions important for sulfate formation not captured by models using prescribed oxidants. Models with prescribed oxidants tend to replenish the oxidants too fast after a complete depletion, e.g., complete titration of H2O2 by SO2 inside clouds. Interactive chemistry shows a stronger oxidation limitation than off-line models. In this article our main focus will be on the coupling between the sulfur chemistry and the oxidants. To test out model parameterizations and validate model performance our model results will be compared with observations. The use of detailed meteorological input data in the OsloCTM2 (liquid water, cloud distribution, precipitation) allows us to study e.g., the impact on the sulfur cycle of chemical interactions in the aqueous phase more thoroughly. Finally we focus on the changes in the anthropogenic emissions from 1985 to 1996 and how these have altered the sulfur cycle. Since meteorological input data represent 1996, emissions and observations for 1996 are chosen for comparison. Many previous studies used the GEIA 1985, therefore we compare sulfur distributions using the 1985 GEIA inventory with our 1996 simulations.

2. OsloCTM2

[5] OsloCTM2 is a global three-dimensional chemical tracer model [Sundet, 1997] using T21 horizontal resolution for this study (5.625° × 5.625°) and 19 vertical layers in hybrid σ coordinates from the surface up to 10 hPa. The time steps are 5 min for chemistry and 60 min for transport. The meteorological input data is generated by the Integrated Forecast System (IFS) from the ECMWF. The IFS model has been run in T63 resolution (1.875° × 1.875°) and truncated to T21. All the flux fields (e.g., convective mass fluxes) are accumulated to T21 grid, while the other grid data are averaged (humidity, cloud properties etc.). The data are sampled either as instant values every third hour or as 3-hour averages. The IFS was run explicitly for 36-hour periods initialized from analysis every 24 hours (12 hours spin-up) to provide a consistent set of input data for the CTM. The IFS model was run for 1996 for this study. Owing to realistic meteorological input data we are able to simulate real weather situations and can perform a comparison between model results and observations for 1996.

[6] The advective transport of tracers is calculated using the Second-Order Moments method [Prather, 1986]. A K-profile scheme from Holtslag et al. [1990] is used to provide eddy diffusion coefficients for the boundary layer mixing. The dry deposition scheme from Isaksen and Rodhe [1978] is applied. Wet deposition by convective precipitation and rain-out in large-scale systems are treated separately based on the meteorological input data (see sections 2.4.2 and 2.4.3). For the chemistry calculations we use the QSSA chemistry solver [Hesstvedt et al., 1978]. The gas phase chemistry scheme is taken from Berntsen and Isaksen [1997] with in all 51 components including 32 NMHC components. This scheme covers the most important parts of the ozone-NOx-hydrocarbon chemistry cycle. Photolysis reaction rates are calculated online using the Fast-J scheme from Wild et al. [2000]. Calculated changes in sulfate aerosols do not affect the photolysis rates.

2.1. Emissions

[7] The meteorological input data represents the year 1996 and it is crucial for us to use an emission inventory for 1996 to make a basis model run to compare with 1996 observations. No source provides a global inventory for 1996 so we had to construct an inventory based on two sources: (1) For Europe and North America we used the total emissions for 1996 as reported by each country available from the UNECE/EMEP emission database WebDab [Vestreng and Klein, 2002]. However, these numbers provide no information about horizontal or vertical distribution inside each country, so we used the horizontal/vertical distribution from GEIA 1985 1B [Benkovitz et al., 1996]. In other words, we scaled the GEIA 1985 1B so that total emissions from each country equal the emissions reported for 1996. (2) For the rest of the world we scale the GEIA 1B emission inventory according to fossil fuel use for each country (fossil fuel use in 1985 versus fossil fuel use in 1996), a method suggested by Karlsdóttir and Isaksen [2000]. Numbers for the fuel consumption are found in BP 1997 Statistical Review of World Energy (June 1998). Anthropogenic emissions of NOx [Benkovitz et al., 1996], CO and VOCs [Olivier et al., 1996] the model are displayed in Figure 1 and the numbers are shown in Table 1. We see that the geographical distribution change from 1985 to 1996 but the total emissions remain almost unchanged. Most pronounced are the decrease in western Europe and the United States, due to cleansing technology and fuel switch, and the increase in Asia, due to economic growth. From 1985 to 1996 we estimate that SO2 emissions decreased by 17.6% in North America and by 47.5% in Europe, whereas it increased by 93% in China and by 31% in Japan. The decrease in eastern Europe and former USSR is a result of economic recession.

Figure 1.

Yearly average SO2 anthropogenic emission inventory for 1985 (upper panel) and 1996 (lower panel) used in the model, unit: 109 molecules cm−2 s−1.

Table 1. Global Anthropogenic Emissions of Sulfur, NOx, and CO Used in the Model for 1985 (1990) and 1996a
 GEIA 19851996
  • a

    References: Benkovitz et al. [1996] concerning GEIA 1985 for sulfur and NOx, Olivier et al. [1996] concerning EDGAR 1990 for CO. Emissions for 1996 are elaborated using EMEP emission inventory for Europe and North America [Vestreng and Klein, 2002] and scaled according to fossil fuel use elsewhere.

  • b

    Five percent of anthropogenic sulfur is emitted as sulfate to account for instant oxidation in the plume. Units: Tg(S).

  • c

    Unit: Tg(N).

  • d

    Unit: Tg.

Sulfur, over 100 m42.9430.57
Sulfur, under 100 m24.0337.20
Totalb66.9766.77
NOxGEIA 19851996
Totalc20.0133.66
COEDGAR 19901996
Totald417.0489.5

[8] EDGAR 3.2 [Olivier and Berdowski, 2001] provides an emission inventory for 1995, but it does not provide vertical distribution of the emissions and their inventory does not show as large decrease in Europe as EMEP does, a decrease that is well documented (and also seen in observations). We choose to put confidence in the EMEP numbers, these are emissions as reported by each country, and thereby take into account the degree of cleansing technology and economic level.

[9] To account for oxidation of SO2 to sulfate in the plume we assume that 5% of anthropogenic SO2 is emitted as sulfate [e.g., Langner and Rodhe, 1991; Pham et al., 1995; Roelofs et al., 1998]). The emissions below 100 m are released partially in model layer 1 (from the surface up to ∼45 m) and partially in model layer 2 (from ∼45 m to ∼190 m) according to the actual thickness of layer 1 [Roelofs et al., 1998]. Emissions above 100 m are entirely released in model layer 2.

[10] SO2 from ships [Endresen et al., 2003] have been included with a total emission of 3.41 Tg(S) year−1. Although this is a small source, ships emit SO2 in areas where oxidation of DMS is the only source of SO2 and may therefore be important. Volcanic emissions of SO2 are taken from Spiro et al. [1992] with vertical distribution from Graf et al. [1997]. Only continuous degassing volcanoes are included and the total is adjusted to 8 Tg(S) year−1 so that 3 Tg(S) is released at the edge of the volcanoes, 3 Tg(S) at the top and 2 Tg(S) is released between 5 and 8 km to account for thermal lifting. Biomass burning emissions of 2.,35 Tg(S) year−1 of SO2 are taken from Spiro et al. [1992]. Emissions of H2S are taken from Spiro et al. [1992] with total emissions of H2S of 0.88 Tg(S) year−1.

[11] Emissions of DMS are calculated based on observed seawater concentrations of DMS [Kettle et al., 1999; Kettle and Andreae, 2000] and 10 m wind from the model input data. The Liss and Merlivat [1986] parameterization is used (hereafter called LM86) where 3 different wave regimes are defined depending on the 10 m wind speed. The oceanic flux of DMS is formulated as

equation image

where Kw is the piston velocity (cm s−1), H is Henry's law constant, Cair is DMS concentration in air and Cocean is DMS concentration in seawater. Cair is normally much lower than the concentration that would be in equilibrium with Cocean so Cair/H is negligible under normal atmospheric conditions. The LM86 parameterization of Kw under different wave regimes is listed in Table 2. Other parameterizations of the DMS emissions are tested in section 3.

Table 2. Piston Velocity (unit cm s−1) Taken From Liss and Merlivat [1986] Used to Calculate Emissions of DMS in the Modela
ParameterValue
  • a

    Scr, Schmidt number for CO2 at 20°T; Sc, Schmidt number for CO2 at sea surface temperature (nondimensional). From Saltzman et al. [1993]: Sc = 2674.0 − 147.12 T + 3.726 T2 − 0.038 T3; T, temperature °C.

Smooth surface regime, ∣U10∣ < 3.6 m s−1Kw = 0.17 × (Scr/Sc)2/3 × ∣U10
Rough surface regime, 3.6 m s−1 ≤ ∣U10∣ < 13.0 m s−1Kw = 2.85 × (Scr/Sc)1/2 × (∣U10∣ − 3.6) + 0.612 × (Scr/Sc)2/3
Breaking wave regime, 13.0 m s−1 < ∣U10Kw = ((5.9 × ∣U10∣) − 49.91) × (Scr/Sc)1/2 + 0.612 × (Scr/Sc)2/3

[12] All other emissions are taken from the GEIA 1985 database (http://geiacenter.org), Müller [1992] and EDGAR emission inventory [Olivier et al., 1996]. For NOx total anthropogenic emissions are 24 Tg (N)/year where we assume that 5% is released as NO2 and 95% as NO.

2.2. O3, OH, and H2O2 Chemistry

[13] A brief discussion of the O3, OH and H2O2 chemistry and distribution is given in the following section. OH and H2O2 are the most important oxidants in daytime chemistry while NO3 are the most important oxidant in nighttime chemistry. All the oxidants are significantly influenced by O3. The full scheme is given in the work of Berntsen and Isaksen [1997]. An overview of some key gas phase reactions in the oxidant chemistry particularly affecting the sulfur cycle is given in Table 3. Formation of tropospheric O3 involves both the NOx cycle and the hydrocarbon cycle through the reactions R1R3. Maximum in tropospheric O3 is found in areas with high emissions of NOx and hydrocarbons and with strong incoming solar radiation. Figure 2 displays surface ozone concentrations for January and July from the basic model run.

Figure 2.

Monthly average O3 in layer 1 (at the ground) for January (upper) and July (lower), unit ppb.

Table 3. Key Gas Phase Reactions in the Oxidant Chemistry Relevant for the Sulfur Cycle
ReactionNumber
equation image
R1
equation image
R2
equation image
R3
equation image
R4
equation image
R5
equation image
R6
equation image
R7
equation image
R8
equation image
R9
equation image
R10
equation image
R11
equation image
R12
equation image
R13
equation image
R14
equation image
R15
equation image
R16
equation image
R17

[14] The only important production pathway for H2O2 is reaction R4. The reaction rate is modified by the presence of water [Kircher and Sander, 1984]. H2O2 is lost by aqueous phase reaction with SO2 and dry and wet deposition in addition to reactions R5 and R6. Reactions 1 and R4 are part of the odd hydrogen cycle, i.e., the rapid cycling between H, OH and HO2. R1 represents the main loss of HO2 for levels of NO > 10 ppt, otherwise R4 becomes the main loss of HO2. Figure 3 shows averaged fields of H2O2 for January and July from the basic model run. In the vertical H2O2 has a maximum in the vicinity of equator with the highest values around 800 hPa (results not shown). In very polluted areas with higher concentrations of SO2 than of H2O2, SO2 may deplete H2O2. It takes some time before H2O2 is replenished after such a titration. In our model this is governed by the photochemistry of the model. In models with prescribed H2O2 concentrations this can be handled either by a relaxation to the prescribed H2O2 concentration with a given time constant [e.g., Seland and Iversen, 1999] or by resetting the H2O2 to the prescribed value every time step [e.g., Chin et al., 1996]. These assumptions could lead to an unrealistic sulfate formation. For instance in the latter case the model tend to overestimate the oxidation of SO2 to sulfate by H2O2 in the aqueous phase. Semiprognostic models [Koch et al., 1999] determine H2O2 based on prescribed OH and HO2 concentrations and J values for H2O2 and will then have a delayed buildup of H2O2 after a titration. Differences among models are described further in section 4.2.

Figure 3.

Monthly average H2O2 in layer 1 (at the ground) for January (left) and July (right), unit ppt.

[15] OH is the most important oxidant in the daytime chemistry. In the troposphere OH is initially formed by the reactions R7 and R8. OH has a maximum at low latitudes due to the high abundance of water vapor and strong incoming solar radiation in this region. Results for OH in the lowermost layer are depicted in Figure 4. Since the main loss of methane occurs via reaction with OH the lifetime of CH4 is often used as a measure for the global OH concentration. In our model the average lifetime of CH4 is 7.8 years. This is in good agreement with other estimates [e.g., IPCC, 2001]. The main loss pathways for OH are oxidation by CH4 and CO, giving HO2 and H respectively although these loss reactions are part of the odd hydrogen cycle and might produce OH back again depending on the local NOx. In polluted regions reaction R1 may enhance OH by shifting the HOx balance toward OH. Figures 24 illustrate that there are large regional variations in the distribution of the oxidants depending on chemical activity. It is therefore important to perform interactive chemical calculations in order to capture the impact of such variations on the chemistry. NO3 is the main oxidant in the nighttime chemistry, it is produced by reactions R9R11. NO3 builds up during night but disappears at dawn due to rapid photodissociation (R12 and R13). HO2NO2 is part of both the HOx- and the NOx-cycle. HO2NO2 is produced by the reaction R14 and is lost by R15R17.

Figure 4.

Monthly average OH in layer 1 (at the ground) for January (left) and July (right), unit 103 molecules cm−3.

2.3. Sulfur Cycle

[16] For this study five new components have been included in the OsloCTM2; DMS, (dimethylsulfide, CH3SCH3), H2S, MSA (methanesulfonic acid, CH3SO3H), SO2 and SO42− (sulfate). H2S and MSA are of minor importance in the sulfur cycle but are included for completeness. Processes included are emissions of DMS, H2S and SO2 (sources) and dry and wet deposition of MSA, SO2 and sulfate (sinks), as well as chemical reactions between these components (gas phase and aqueous phase). Mixing ratio of sulfate and MSA are calculated without any treatment of particle size or size distribution. An overview of all the processes is given in Figure 5.

Figure 5.

The sulfur cycle included in the OsloCTM2 with emissions, deposition and oxidation pathways; for budget numbers, see Table 7.

2.3.1. Gas Phase Oxidation

[17] Gas phase oxidation of sulfur compounds is given in Table 4. DMS is oxidized by NO3 (reaction R18) and by OH (reactions R19 and R20). NO3 is an important oxidant at nighttime. Oxidation of DMS is a complex process [see, e.g., Turnipseed and Ravishankara, 1993]. In order to simplify the calculations we have omitted the intermediary products in the oxidation chain and assume that SO2 and MSA are formed directly from DMS. MSA is then lost by dry and wet deposition. H2S is oxidized to SO2 in the gas phase by OH (reaction R21). Gas phase oxidation of SO2 by OH forms sulfate (reaction R22).

Table 4. Gas Phase Reaction Used in the Sulfur Schemea
ReactionReaction RateNumber
equation image
k = 1.9 × 10−13 × exp(520/T)R18
equation image
k = 1.2 × 10−11 × exp(−260/T)R19b
equation image
k = ([O2] × 1.7 × 10−42 × exp(7810/T))/(1 + ([O2] × 5.5 × 10−31 × exp(7460/T)))R20c
equation image
k = 6.0 × 10−12 × exp(−75/T)R21
equation image
kOH = (k0/(1 + k0/equation image)) × equation image k0 = 3.0 × 10−31 × (300/T)3,3 × [M] equation image = 1.5 × 10−12R22

2.3.2. Aqueous Phase Oxidation

[18] SO2 is also oxidized to sulfate (S(iv) → S(vi)) in the aqueous phase by H2O2, HO2NO2 and O3 (Table 5). Oxidation of SO2 by HO2NO2 in the aqueous phase is of minor importance except for some remote areas. To simulate the aqueous phase chemistry a heterogeneous scheme is included [Jonson and Isaksen, 1993; Jonson et al., 2000]. The main idea is to calculate the fraction of a gas that is dissolved in the cloud droplets and modify the reaction rates. Another method would be to calculate gas phase and aqueous phase separately and then combine. To modify the reaction rates makes this scheme easy to include in the gas phase scheme already used in the model. It is also computational efficient. For this we need the following parameters; Henry's law constants, equilibrium rates and reaction rates (listed in Table 5). We calculate modified reaction rates for reaction of total SO2, including aqueous phase oxidation by H2O2, HO2NO2 and O3, for reaction R5 (Table 3) we take into account the fraction of SO2 in the gas phase. To account for the solubility and dissociation of SO2 in the aqueous phase we apply an effective Henry's law constant [see Seinfeld and Pandis, 1998, pp. 346–347]. We assume a pH of 4.5 inside clouds, in accordance with, e.g., Koch et al. [1999].

Table 5. Henry's Law Constants, Aqueous Phase Reaction Rates, and Effective Henry's Law Constant Applied in the Aqueous Chemistry Scheme and in the Wet Deposition Scheme
EquilibriumEquilibrium Rate
Henry's Law Constants
O3(g) ↔ O3(aq)H(O3) = 1.13 × 10−2 × exp(2300 × Tfac)a
HNO3(g) ↔ HNO3(aq)H(HNO3) = equation imageb
CH2O(g) ↔ CH2O(aq)H(CH2O) = 3.2 × 103 exp(6800)c
H2O2(g) ↔ H2O2(aq)H(H2O2) = 7.1 × 104 × exp(6800 × Tfac)d
HO2NO2(g) ↔ HO2NO2(aq)H(HO2NO2) = 1.2 × 104 × exp(6900 × Tfac)e
SO2(g) ↔ SO2(aq)H(SO2) = 1.23 × exp(3020 × Tfac)f
MSA(g) ↔ MSA(aq)H(MSA) = 5.0 × 104 (mol/atm.)g
ReactionReaction Rate
Aqueous Phase Reaction Rates
SO2(aq) ↔ HSO3(aq) + H+(aq)K′ = 1.23 × 10−2 × exp(2.01 × 103 × Tfac)f
HSO3(aq) ↔ SO32−(aq) + H+(aq)K″ = 6.0 × 10−8 × exp(1.12 × 103 × Tfac)h
H2O2(aq) + HSO3(aq) ↔ H+(aq) + SO42−(aq) + H2OkH2O2 = (8.0 × 104 × exp(−3650 × Tfac))/(0.1 + [H+])d,i
O3(aq) + SO32−(aq) ↔ SO42−(aq) + O2(aq)kO3 = 1.8 × 104 × [H+]−0.4i,j
HO2NO2(aq) + HSO3(aq) ↔ 2H+(aq) + SO42−(aq) + NO3kHO2NO2 = 3.1 × 105i,k
 
Effective Henry's Law Constant
SO2(g) ↔ SO2(aq)Heff(SO2) = H(SO2) × [1 + K′/[H+] + K′K″/[H+]2]l

[19] We define a fraction of a tracer within the cloud which is dissolved in a cloud droplet, faq:

equation image

Hc is Henry's law constants, T temperature, R universal gas constant and clw is volume fraction of liquid water within the cloud (clw is divided by the local cloud fraction). For SO2 we use the effective Henry's law constant. HNO3 and sulfate are so soluble, i.e., Henry's law constants are so large, that we can assume that everything is dissolved in the droplets without using equation (2).

[20] The R22 gas phase reaction rate kOH is thus modified

equation image

where cl is fractional cloud cover, fSO2 designate faq for SO2. The term (1 − fSO2). cl represents gas phase inside the cloud and (1 − cl) represents cloud free part of the grid box.

[21] In the aqueous phase the production rate of sulfuric acid is defined as:

equation image

where H2O2, SO2, O3, H+, HSO3, and HO2NO2 represent the concentrations of these species respectively.

[22] Then the aqueous phase reaction rates are modified so that

equation image
equation image
equation image

where Ψ is a conversion factor, Ψ = 103/(A0 × clw) to convert the aqueous phase rates from L mol−1 s−1 to cm3 molecules−1 s−1 (103: cm3/L, A0: Avogadro's number, clw: volume fraction of liquid water). fO3, fH2O2, and fHO2NO2 represent faq for these gases.

[23] Hence total production of sulfate from SO2 (gas and aqueous phase) is:

equation image

SO2(tot) is concentration of SO2 in the gas and aqueous phase together (molecules/cm3). The rates k′OH, k′H2O2, k′O3 and k′HO2NO2 (equations (3) and (5)(7)) are used in the QSSA chemistry solver [Hesstvedt et al., 1978] along with all other gas phase rates included in the model. This method assumes an infinite exchange of species between clouds and cloud-free air and thus that the grid box is in equilibrium whereas the atmosphere is not in equilibrium. The error introduced with this assumption gets larger with larger grid sizes. The oxidation of SO2 by O3 inside clouds is very sensitive to pH. We assume a fixed pH = 4.5, since calculations of cloud pH require components and ions not included in the model. In some regions (e.g., China) compounds such as mineral dust affect atmospheric pH significantly, and even an ion balance model would have difficulties to represent pH properly there. From the expression of kO3 in Table 5 we see that the rate decreases by a factor 1.58 if pH decreases with a value 0.5. A sensitivity test has been performed with a fixed pH of 4 (see section 4.1.2).

[24] To account for oxidation of SO2 to sulfate by metal ions in the aqueous phase, mainly manganese Mn(II) and iron Fe(III), a minimum loss rate of kcat = 2.78 × 10−6 s−1 is introduced [Seland and Iversen, 1999]. This loss rate corresponds to a lifetime of 100 hrs. North of 45°N a loss rate kcat = 5.56 × 10−6 s−1 (50 hours lifetime) is applied assuming that the Northern Hemisphere (hereafter denoted NH) is more polluted. This loss rate is scaled according to cloud fraction.

[25] A test using monthly averages of the oxidants taken from a model run without the sulfur cycle is performed to see how the online calculation of the oxidants (i.e., with oxidation limitation) affect the sulfur cycle (results not shown). The effect is largest in heavily polluted areas in NH, and particularly in January where we experience an oxidant limitation using online oxidants. With prescribed oxidants the transfer of SO2 to sulfate proceeds faster and gives 31% less global mass of SO2 and 4% more global mass of sulfate in January than in the run with online oxidants. In July global mass of SO2 decreases with 8% and global sulfate increases with 1%, i.e., there is less oxidation limitation. Maximum impact of fixed oxidants is found in China, where maximum sulfate in January increases with 120% using prescribed oxidants. SO2 is most likely to deplete oxidants in areas with high SO2 concentrations (like China). The consequence of this is that the effect of the oxidation limitation is largest in this region. For further analysis of off-line versus online see section 4.2.

2.4. Model Deposition Processes

2.4.1. Dry Deposition

[26] Dry deposition scheme from Isaksen and Rodhe [1978] expanded in the work of Berntsen and Isaksen [1997] is included in the model. For all tracers undergoing dry deposition there are one winter and one summer value for the five vegetation types water, forest, grass, tundra/desert and ice/snow. These seasonal values are scaled according to the fraction of each vegetation type given in the model. Dry deposition of MSA, SO2 and SO42− is done in the same manner by using the velocities given in Table 6. Together with the actual thickness of layer 1 we get a dry deposition loss term that is then used in the chemistry routine using 5 min time step. We assume snow covered ground for temperatures below −5°C (snow cover is not included in the ECMWF meteorological data used in the OsloCTM2).

Table 6. Dry Deposition Velocities Used in the Modela
 MSAbSO2cSO42−c
Ocean0.20.80.2
Land0.20.60.2
Ice/snow0.20.10.2

2.4.2. Wet Deposition in Large-Scale Systems

[27] All soluble tracers are removed in large-scale cloud systems according to the in cloud scavenging scheme by Berge [1993]. The in-cloud scavenging rate is calculated according to the formula:

equation image

(unit: s−1) where pr is precipitation (kgwater m−2 s−1) released from a grid box, Δz is the height (in m) of the grid box, CLW is cloud water (kgwater m−3), calculated from the cloud water content and the air density and grid volume. faq is the fraction of a specie inside the cloud dissolved in cloud water defined in section 2.3. For HNO3 and SO42− we assume that all mass inside a cloud is dissolved in the cloud droplets (faq = 1). Gases are partially released in case of evaporation of falling rain, but HNO3 and SO42− are not released unless all rain is evaporated. Wet deposition in ice clouds is not considered although laboratory experiments by Conklin et al. [1993] indicate that there is some deposition of SO2 onto ice cloud particles.

[28] Another aspect concerning these input data is that they apply for the entire grid box while in the real atmosphere there will be areas with rain and wet deposition and areas without. As an example: if 70% of the water in a grid box is lost by rain this does not imply that 70% of sulfate is lost as well since the rain could be limited to a certain cloud fraction of the grid box. If we do not take this into account we may overestimate the wet deposition. We have introduced a constraint so that we cannot remove more than a certain fraction during a 3-hour period. This maximum fraction is determined by the cloud fraction and by the fraction of the grid box that is exchanged due to advection. We have introduced this advection term since the wind may blow through the clouds and hence make more air available for cloud processes. For example, if the cloud fraction is 20% and 4% of the air is exchanged due to advection we know that at most 24% of the air in that grid box will be in contact with cloud air during a 3-hour time step. Then we adjust the wet deposition so that at most 24% of the tracer is removed by wet deposition in this 3-hour period. This constraint defines an upper limit.

2.4.3. Wet Deposition in Convective Events

[29] The mass fluxes due to convection in the OsloCTM2 are diagnosed as updrafts, downdrafts, updrafts entrainment and downdrafts entrainment (kg m−2 s−1). In the convection parameterization, mass is redistributed in the vertical by a so-called “elevator.” The idea is in short that if entrainment is larger than detrainment mass is put into the “elevator” and moved upward. If detrainment is larger than entrainment, some of the surplus in the “elevator” is redistributed to the actual grid box. During convection a certain fraction of the mass in the elevator is removed by wet deposition. This fraction is dependent upon the fraction of the specie dissolved in the cloud droplets. For HNO3, SO42−and MSA everything is dissolved in the cloud droplets (faq = 1). The input cloud parameters apply to the entire grid box but here we are interested in only the convective part, it is crucial to determine how much of the water we put into the elevator is actually condensed and how much of the water remains in gas phase. For this purpose an iterative method is applied described further in Appendix A. Both wet deposition in large-scale systems and in convective events are part of the transport and hence use 60 min time steps.

2.4.4. Subcloud Deposition

[30] SO2 is scavenged according to the rates determined by Martin [1984]:

equation image

where rainfall is given in mm rain hour−1 or kgwater m−2 hour−1. These rates are empirically determined based on field measurements.

[31] Sulfate is scavenged according to Berge [1993]:

equation image

C = 5.2 m3 kg−1 s−1, mp is precipitation mass (kgwater m−3) and Em is a mean collection efficiency. We have applied the value Em = 0.1 from Tremblay and Leighton [1986]. Qsubcloud has unit s−1 and is used in the chemistry routine (time step 5 min). Both SO2 and sulfate subcloud scavenging rates are scaled according to the fraction of the grid box that contain falling water. In a raining column we find the maximum cloud fraction up to approximately 4000 m, we assume that the cloud base is never located above 4000 m. Then we take 20% of this maximum cloud fraction, i.e., we assume that 20% of the cloud is precipitating. From Table 7 we see that subcloud scavenging of SO2 represents about equation image of the total wet deposition of SO2 (1.3% of total loss). For sulfate, the subcloud scavenging is negligible compared to large-scale precipitation and convective precipitation. Many models have omitted subcloud scavenging of sulfate due to its minor importance.

Table 7. Calculated Budget (Sources and Sinks) for the Sulfur Species Using 1996 Emission Inventory in the Model
 Absolute Number of Sources and Sinks, Tg(S)Percentage
DMS
Oceanic emissions11.95 
 
Sinks
NO3 oxidation3.2727.4
OH oxidation (abstraction)4.4136.9
OH oxidation (addition)4.2735.7
Total sinks11.95 
 
SO2
Anthropogenic emissions <100 m29.0432.3
Anthropogenic emissions >100 m35.3439.4
Ship emissions3.413.8
Volcanic emissions8.008.9
Biomass burning emissions2.252.5
DMS oxidation10.8812.1
H2S oxidation0.881.0
Total sources89.80 
 
Sinks
Wet deposition (large scale)0.210.2
Convection0.200.2
Subcloud scavenging1.161.3
Dry deposition41.4946.2
Oxidation by OH (gas)7.948.8
Oxidation by O3 (aqueous)5.806.5
Oxidation by H2O2 (aqueous)28.9032.2
Oxidation by HO2NO2 (aqueous)1.601.8
Catalytic (metals)2.412.7
Stratosphere0.090.1
Total sinks89.80 
 
SO42−
Anthropogenic emissions <100 m1.533.1
Anthropogenic emissions >100 m1.863.7
SO2 oxidation46.6593.2
Total sources50.04 
 
Sinks
Wet deposition (large scale)39.4078.7
Convection3.226.4
Subcloud scavenging0.030.1
Dry deposition7.3914.8
Total sinks50.04 

3. Model Results

3.1. Year 1996 Sulfur Distribution and Budget

[32] Model results for DMS, SO2 and sulfate in the lowest model layer and zonal averages are shown for January and July (Figures 6, 7, and 8). There are large regional and seasonal differences, particularly in the DMS and SO2 distributions. In January a maximum of 4.75 ppbv is found for DMS in the Southern Hemisphere (hereafter denoted SH) near the Antarctic ice shelf (Figure 6), while in July a maximum of 0.95 ppbv is found at 60°N, in other words DMS maxima are found in the summer hemispheres where emissions are strongest. Ocean concentrations of DMS show a large maximum in December/January in the South Pacific just north of the Antarctic ice shelf [Kettle et al., 1999]. Using the LM86 parameterization of the sea/air flux this gives the estimated large maximum in the DMS concentrations shown in the marine boundary layer near Antarctica in December and January. Over the ocean south of 60°S all values exceed 400 pptv in January. Even though there are relatively high abundances of oxidants in the summer hemisphere the emissions are sufficiently strong to increase the concentrations substantially. Zonal average of DMS shows that little DMS is transported out of the boundary layer, except around 50°S to 70°S.

Figure 6.

(top) Monthly average DMS at the ground and zonal (bottom) monthly average for (left) January and (right) July, unit ppt.

Figure 7.

(top) Monthly average SO2 at the ground and (bottom) zonal monthly average for (left) January and (right) July from a model run using the 1996 anthropogenic emission inventory, unit ppt.

Figure 8.

(top) Monthly average sulfate at the ground and (bottom) zonal monthly average for (left) January and (right) July from a model run using the 1996 anthropogenic emission inventory, unit ppt.

[33] Figure 7 shows the effect of the industrialized emissions in the NH; United States, Europe and Southeast Asia on the distribution of SO2. Maximum values at the ground exceed 25 ppbv and 11 ppbv for January and July respectively. Maximum concentrations are found in SE Asia. However, the overall maximum concentration is found in model layer 2 (∼45–180 m) because a large part of the anthropogenic sulfur emissions occurs above 100 m. The January maximum in layer 2 is about 30 ppbv and the July maximum is 17 ppbv. The high concentrations in January are due to a combination of high emissions in NH winter coupled with the oxidation limitation and a shallow boundary layer. This also implies that more anthropogenic SO2 is lost by dry deposition directly after release in January than in July. Observations also show high concentrations in NH winter. Our model gives high values of SO2 in the lowermost kilometer, which corresponds to the boundary layer, although some SO2 is transported upward by convection at NH midlatitudes.

[34] Maximum values of sulfate (Figure 8) are found in Southeast Asia where emissions are high and where the oxidation is efficient. Regions of high sulfate are related to areas of high SO2 and down-wind of those. Similar to SO2 the maximum concentration is found in layer 2. The January/July maxima of sulfate in layer 1 are found over China with 5.2 ppbv and 4.5 ppbv, respectively (layer 2: 5.6/7.2 ppbv). A distinct difference between SO2 and sulfate loss is that a large fraction of SO2 is lost shortly after emission through dry deposition whereas the sulfate loss by wet deposition occurs over a larger area. In the vertical we find an enlarged maximum of sulfate which goes above the boundary layer, we also see that sulfate is transported upward by convection. In addition sulfate also displays a secondary maximum in the lower stratosphere at NH midlatitudes caused by transport of anthropogenic sulfur from the troposphere (more discussed in the work of Myhre et al. [2004]). If sulfate is transported upward by convection and if it is not washed out it may eventually reach the stratosphere where there is no loss process of sulfate other than the deposition velocity (2.5 cm hour−1). Hence sulfate has a long lifetime and a maximum in concentration in the stratosphere. Total columns of sulfate (up to 10 hPa in the model) for January and July are shown in Figure 9.

Figure 9.

Monthly average total column of sulfate for (left) January and (right) July from a model run using the 1996 anthropogenic emission inventory, unit: μg(SO42−) m−2.

[35] A global yearly budget for the sulfur cycle is listed in Table 7. About 75% of the emissions of sulfur are anthropogenic. Of the remaining 25%, half is released as DMS and half is released as volcanic and biogenic SO2. This illustrates the dominating anthropogenic contribution to the sulfur cycle. Wet deposition and dry deposition account for ∼50% each of total loss. Regionally the DMS emissions are most important in remote background areas over the oceans while SO2 and sulfate are most important in the industrialized areas in the NH. Yearly average burden and lifetime for the sulfur species are listed in Table 8 and the results are further discussed in section 4.

Table 8. Burden and Lifetime (Turnover Time, i.e., Total Loss Compared to Average Mass) for the Three Main Sulfur Components in the Modela
 DMSSO2: 1996(1985)SO42−: 1996(1985)
  • a

    Values for model run using the GEIA 1985 anthropogenic emission inventory are in parentheses. Concerning DMS the differences in burden and lifetime using the 1996 versus GEIA 1985 emission inventories are so small that it can be ignored.

Burden, Tg(S)0.0630.25(0.26)0.53(0.50)
Lifetime, days1.930.99(1.06)3.84(3.69)

3.2. Comparison With Observations

[36] Comparing model results with observations is not a straightforward task although it is the only way to validate model performance. The distribution of SO2 and sulfate are very dependent upon the emissions of SO2 and the wet removal processes. The emissions and meteorological conditions vary from one year to another and hence the levels of SO2 and sulfate will vary as well. This will introduce an error if the model data is not representative for the period of the observations. Another important aspect is the model resolution since observation sites may be influenced by local conditions not captured by the much coarser model grid. In these comparisons we focus on two areas: (1) a remote background area in the SH where the sulfur budget is mainly affected by natural DMS emissions and (2) a NH region dominated by anthropogenic SO2 emissions.

3.2.1. Southern Hemisphere Background Oceanic Region

[37] There exist few observations in SH remote regions and to make the situation even worse 15 SH and Pacific sites were closed down in 1996 [Barrie et al., 2001]. In this article we stress the fact that we use the year 1996, to validate model performance we compare with results from the PEM-Tropics-A performed from 15 August to 15 October 1996 [Emmons et al., 2000]. Figures 10 and 11 show model results of DMS and SO2 (for 15 August to 15 October) compared to observations for the five regions covered by PEM-Tropics-A.

Figure 10.

Comparison of DMS model results versus observations for the five regions covered by the PEM-Tropics-A campaign from 15 August to 15 October 1996 [Emmons et al., 2000], unit ppt.

Figure 11.

Comparison of SO2 model results versus observations for the five regions covered by the PEM-Tropics-A campaign from 15 August to 15 October 1996 [Emmons et al., 2000], unit ppt.

[38] For DMS (Figure 10) the model reproduces the vertical pattern fairly well with high values near the surface (∼50–70 ppt) and rapidly decreasing values with height. The model tends to underestimate DMS around 1.5–3 km, this may be due to the formulation of the boundary layer height in the model (too shallow boundary layer). Concerning SO2 (Figure 11) the model underestimates the observations. In four of the regions (not Fiji) the model gives a vertical pattern with high values in the boundary layer rapidly decreasing with height. The observations decrease with height but show no sharp boundary layer gradient, except for Fiji where our model does not. In some areas, e.g., Hawaii, the model results have an increase in the middle troposphere due to emissions from volcanoes. It seems that our model corresponds better with the observations from the DC8 flights than the P3 flights, although we have no explanation for this. Emmons et al. [2000] also report sulfate from PEM-Tropics-A, but that is sulfate in the gas phase (typical values 106 molecules/cm3), while we have total sulfate in our model (typical values 108–109 molecules/cm3) so no comparison is possible.

[39] As shown in Figure 6 the model gives a maximum of 4.75 ppbv of DMS near Antarctica in January. Searching through the literature we have not found any observations of DMS as high as this, not for 1996, nor for any other year. The model clearly overestimates DMS in this region in December and January. On the other hand, using the Kettle et al. [1999] seawater concentrations and the LM86 parameterization we calculate emissions of 11.95 Tg(S) DMS per year (as did, e.g., Koch et al. [1999]). This is at the lower end of the emissions reported in the literature, and may be too low. Different parameterizations of the DMS flux give substantial different emissions even when all use the Kettle et al. [1999] oceanic inventory [Boucher et al., 2003]. We have tested other parameterizations [Wanninkhof, 1992; Nightingale et al., 2000], and both give higher global emissions of DMS. At the same time these other parameterizations give very high concentrations in December/January near Antarctica. Using Wanninkhof [1992], where the emissions are proportional to the ten meter wind quadrated, we get a flux of 26 Tg(S) year−1 of DMS and a maximum of 8 ppbv in January. Even though these other parameterizations with higher emissions may give better correspondence with observations elsewhere we feel that the erroneous high concentrations calculated for January in Antarctica should imply that we keep the LM86 parameterization. Another explanation for these high values may be that the Kettle et al. [1999] concentrations are too high, i.e., that a few single observations of high oceanic concentrations have been made valid for a larger region. Slow boundary layer ventilation may also influence the model results. For example in December in the grid box with the highest model value of DMS the values drop from 5.8 ppb near the surface to 0.3 ppb at 1.2 km. Yet another explanation may be that there are other oxidation pathways for DMS yielding SO2 not included in the model. Such missing pathways other than OH and NO3 could be gas and aqueous phase oxidation by O3 or oxidation by BrO [Boucher et al., 2003].

3.2.2. EMEP Region

[40] The EMEP program also provides observations for Europe. In 1996 85 and 80 stations provided observations for SO2 and sulfate respectively [Hjellbrekke and Hanssen, 1998a, 1998b]. Data for all stations during summer (JJA) and winter (DJF) seasons are summarized in Figure 12 and compared with model estimates. Results comparing monthly averages from the model with observations are shown in Figures 13 (SO2) and 14 (sulfate). Model results using GEIA 1985 emission inventory are also included as well as observations for 1985. All stations within a model grid box are averaged to compare with the model. From Figure 12 we see that the model overestimates SO2 and underestimates sulfate in wintertime (DJF), in summer SO2 is a bit high but sulfate corresponds well with observations. Figures 13 and 14 show that in general the model run using 1996 emission inventory and the 1996 observations show smaller values than the 1985 case (run with 1985 emission inventory and observations). The run using 1996 emissions correspond better with observations from 1996 than from 1985, and the run using 1985 inventory corresponds best with observations from 1985 (meteorological data are the same in the two runs). Some discrepancies exist, especially among the northernmost stations. In the real atmosphere SO2 has maximum values and sulfate minimum values in winter, this seasonal pattern is reproduced by the model. Most studies on the sulfur cycle [see, e.g., Roelofs et al., 2001] overestimate SO2 and underestimate sulfate at high latitudes during winter. Three phenomena contribute to this: high emissions, a shallow boundary layer and low abundance of oxidants. The emissions of SO2 are larger in winter than in summer due to increased coal burning for heating and electric power, NH emissions of SO2 are 15% higher in January than in July. Inversions often occur, this will trap the emissions near the surface. On the basis of the comparison of observations with modeled SO2 it seems that the parameterization in the IFS model gives a too shallow boundary layer over NH continents (North America, Europe) in winter. Less available oxidants in wintertime (especially OH and H2O2) leads to a slower oxidation of SO2, a so-called oxidation limitation: OH depends directly upon sunlight (reactions R7 and R8 (Table 3), while H2O2 depends on the levels of HO2 (reaction R4) which in turn depends on OH and the levels of hydrocarbons.

Figure 12.

Scatterplot of the model results versus observations in Europe (EMEP) for 1996 summer (top; June, July, August) and winter season (bottom; December, January, February), (left) SO2 and (right) sulfate. Unit: μg(S) m−3.

Figure 13.

Comparison of SO2 model results with observations for 16 grid boxes in the EMEP area. Thick line, model result using 1996 emissions; dotted line, model result using GEIA 1985 emission inventory; dots, observations for 1996; stars, observations 1985. Unit: μg(S) m−3. For the observations, an average of all stations inside a grid box is reported. The station codes and names refer to the stations in that particular grid box that reported observations for 1996. For 1985, observations are reported for the stations available.

Figure 14.

Same as Figure 13, but for sulfate. Note that the scale used for sulfate is half of the scale used in the corresponding grid box for SO2.

[41] Chin et al. [2000] suggest that underestimation of sulfate may be due to the fact that the EMEP data are not corrected for sea spray, sea-salt contribution is assumed to be largest in winter because of strong winds. This may partly explain the discrepancy in sulfate, but can not explain the high values of SO2. Another likely possibility is that there are oxidation pathways in the atmosphere that are not included in the model. We have introduced a parameterized catalytic pathway to account for oxidation by metals in the aqueous phase, although these processes are poorly known. These pathways account for over 10% of SO2 oxidation in NH winter, but are not important globally. However, based on the results presented here we believe that our model gives a more sophisticated and more realistic representation of the regional and global distribution of sulfur than models using prescribed oxidants, and that we can attribute the reduced oxidation of SO2 to sulfate in wintertime to lacking oxidation pathways active in periods with low oxidation through known pathways.

[42] The EMEP observation network covers most of Europe and the weather regime differs considerably within this area, from near tropic conditions in the south up to the Arctic. The stations Braganca (PT 1) and Toledo (ES 1) has a winter maximum of SO2, but are influenced by the high pressure system over the Azores moving northward in summer giving a maximum in observed SO2 also in August. Sulfate has a maximum in summer.

[43] A day-to-day comparison of model results and observations (not shown) shows that the model is able to reproduce the observations, both the overall level and specific episodes. Again the model has a tendency to overestimate SO2 in winter, but for sulfate the model result agree well with observations. For Jungfraujoch (3573 m altitude) model results in the grid box corresponding to the actual height of the station agree well with observations. This shows that Jungfraujoch is located in the free troposphere and is not affected by boundary layer processes.

3.2.3. North America

[44] SO2 and sulfate from 1996 for 8 stations in the IMPROVE network in North America are shown in Figures 15 and 16. These stations are both background stations and stations affected by anthropogenic emissions. In general we see that the model overestimates SO2 while sulfate is reproduced reasonably well except for the northern station (Denali) where both SO2 and sulfate is overestimated. The model gives a seasonal pattern of SO2 with high values in autumn/winter and low values in late spring/summer. This pattern is seen at some stations, e.g., Shenandoah NP (VA, this stations lack observations for October to December), while other stations have maximum in observed SO2 in summer. Sulfate is well reproduced, both concerning overall level and seasonal variation. For Denali (Alaska) there are two possible explanations for the overestimation, one is that the model may have some problems with the transport close to the poles and across the poles. Another likely explanation is that the station is located downstream of the Alaska Range and hence that there has been a considerable rainout of sulfate and that the air that reaches the station is relatively clean. This will not be captured by the model. Acadia National Park (Maine, 44.4°N) is located right on the border between two grid boxes N/S, in Figure 15 and 16 both grid boxes are plotted. For the two southernmost stations, Okefenokee Nwr (GA) and Chassahowitzka (FL), we see that SO2 is overestimated and sulfate underestimated in winter. There may be several possible reasons for these discrepancies between model and observations; an oxidation limitation in winter (Okefenokee Nwr and Chassahowitzka), or that the model representation of the mixing in the boundary layer and the transport out of the boundary layer may not be well represented, this will affect SO2 more since it is a primary compound. Another possible explanation is that the T21 grid may be too coarse to represent local phenomena.

Figure 15.

SO2 for eight stations in the IMPROVE network taken from the NAtChem database (Environment Canada, year 1996, Canadian National Atmospheric Chemistry (NATChem) database, Meteorological Service of Canada, 4905 Dufferin Street, Toronto, Ontario, Canada, M3H 5T4). Unit: μg m−3.

Figure 16.

SO42− for eight stations in the IMPROVE network taken from the NAtChem database (see Figure 15 for reference). Unit: μg m−3.

3.3. Hemispheric Differences in Sulfur Processes

[45] The NH sulfur cycle is dominated anthropogenic areas while SH is dominated by oceans and emissions of DMS. To reveal the hemispheric differences we have made budgets for two regions, one midlatitude polluted region located in Europe (39°N to 61°N, 8°W to 31°E) and one background region located in the Pacific (61°S to 39°S, 154°W to 115°W), see Table 9. The regions are of equal size and at same distance from equator. The European region is mostly land covered while the background Pacific region is located over the ocean. For comparison the global numbers are also given.

Table 9. Relative Importance (%) of the SO2 Loss Processes and Correspondent Turnover Time in Days for Two Different Regimes, One Background Region (Pacific) and One Polluted Anthropogenic Region (Europe)a
 Pacific, SummerEurope, WinterGlobal
%τ (d)%τ (d)%τ (d)
  • a

    Results for summer and winter season are shown. Anthropogenic emission inventory for 1996 is applied.

Process of January 1996
Wet deposition2.0 2.5 2.4 
Dry deposition25.53.468.22.050.12.5
OH oxidation, gas3.624.31.499.85.622.2
H2O2 oxidation, aqueous60.01.46.222.125.64.8
O3 oxidation, aqueous4.320.09.314.99.113.6
HO2NO2 oxidation, aqueous0.51892.262.62.157.9
Stratosphere0.18850.0-0.0-
Catalytic4.021.510.213.55.124.5
Total100.00.9100.01.4100.01.2
 Pacific, WinterEurope, SummerGlobal
%τ (d)%τ (d)%τ (d)
Process of July 1996
Wet deposition5.8 1.0 1.4 
Dry deposition17.310.954.01.642.01.9
OH oxidation, gas1.413313.66.211.86.9
H2O2 oxidation, aqueous31.76.026.33.237.42.2
O3 oxidation, aqueous31.76.02.830.24.717.4
HO2NO2 oxidation, aqueous3.554.01.366.31.651.9
Stratosphere0.29650.111000.2330
Catalytic8.422.50.993.00.986.0
Total100.01.9100.00.8100.00.8

3.3.1. Northern Anthropogenic Region

[46] Overall dry deposition dominates as loss of SO2 in both seasons. In winter there is a combination of three phenomena that favors dry deposition: high emissions, high frequent inversions that trap the emissions in the boundary layer and very low abundance of oxidants as discussed previously. Our fraction lost by dry deposition is larger than most other models, in the COSAM exercise, Roelofs et al. [2001] find an average of 50% in winter and 30% in summer in NH anthropogenic regions for all the models. Even if more SO2 is lost by dry deposition in winter than in summer the lifetime with respect to dry deposition is longer in winter since the dry deposition velocity over ice and snow is smaller than over snow- and ice-free surfaces (0.1 cm s−1 versus 0.6 cm−1). The change in oxidation pathways from summer to winter with low abundance of oxidants in wintertime is the main reason for the distinct difference in lifetime: 1.4 days in January versus 0.8 days in July. O3 is as important as H2O2 in winter.

3.3.2. Southern Background Region

[47] Aqueous phase oxidation dominates both in SH summer (January) and winter (July), with more than 70% of the loss. In winter loss by H2O2 and O3 are of equal size (about 30% each) while in summer the loss by H2O2 dominates and oxidation by O3 is insignificant. This is due to seasonal differences of the two oxidants. H2O2 has a maximum in summer (January) whereas O3 has a minimum in summer (10–15 ppb in January) and higher in winter (25–30 ppb). Dry deposition is more important in summer than in winter. This is because most of the SO2 in the remote atmosphere originates from DMS oxidation. In summertime the levels of oxidants are high so that the large emissions of DMS are rapidly oxidized to SO2 in the boundary layer with higher probability that the SO2 will undergo dry deposition.

4. GEIA 1985 Emissions

4.1. Effects of Emission Changes From 1985 to 1996

[48] Most of the studies on the sulfur cycle published over the last decade use the GEIA 1985 anthropogenic emission inventory. Although we focus on 1996 we have performed a simulation with the GEIA inventory to study the impact of the emission changes from 1985 to 1996 and to compare with previous works. Meteorological data for 1996 are used in this run. Emissions of DMS are the same in 1985 and 1996. Figures 17 and 18 show results of SO2 and sulfate for January and July using GEIA 1985 emission inventory.

Figure 17.

Monthly average SO2 in layer 1 (at the ground) for (left) January and (right) July from a model run using the GEIA 1985 anthropogenic emission inventory, unit ppt.

Figure 18.

Monthly average sulfate in layer 1 (at the ground) for (left) January and (right) July from a model run using the GEIA 1985 anthropogenic emission inventory, unit ppt.

[49] The global anthropogenic sulfur emissions are virtually unchanged from 1985 (66.97 Tg(S)) to 1996 (67.77 Tg(S)). However, the regional pattern changed significantly. The changes in emissions from 1985 to 1996 do not alter the overall lifetimes or the sulfur budgets by more than a few percent (see Table 8). However, there are large regional changes, especially in Europe and Southeast Asia. To study how these changes in emissions have altered the sulfur cycle regionally we first analyze the zonally averaged sources and sink of SO2. We will then study the regional changes in the loss processes in Europe and Southeast Asia. Finally we will study the changes in the OH oxidation from 1985 to 1996, since oxidation by OH will contribute to the formation of new sulfate aerosol particles.

4.1.1. Latitudinal Changes in Loss Processes

[50] Although there was a longitudinal shift in the emissions from 1985 to 1996 the latitudinal shift (see Figure 1) implies a change in the efficiency of the oxidation processes. Figure 19 shows a latitudinal plot of the sources of SO2, loss of SO2 and difference between sources and loss (i.e., net production) for both 1985 and 1996. Only the anthropogenic sources of SO2 differ between 1985 and 1996. If we compare the sources and sinks for one of the years we see that they are very similar, both in pattern and magnitude, due to the short lifetime of SO2. However, we have a southward shift from 1985 to 1996; from maximum values located between 30°N and 60°N in 1985 to maximum values located between 15°N and 50°N in 1996. The lower panel shows the net SO2 production for 1985 and 1996 (sources19xx ÷ loss19xx). The region north of 55°N is a net sink of SO2 due to northward transport from Europe. The net loss here is reduced from 1985 to 1996, the emissions are reduced but the sinks are reduced even more. The largest changes are seen between 25°N and 55°N, the regions of net production are moved southward. In the tropics and SH there are minor changes only.

Figure 19.

(top) Total sources of SO2 (emissions and oxidation of DMS and H2S) for 1985 and 1996, (middle) total loss of SO2 for 1985 and 1996, and (bottom) the difference between total sources and loss (i.e., net production) for 1985 and 1996, unit Tg(S) year−1.

4.1.2. Regional Changes in Loss Processes

[51] To demonstrate how the loss processes have changed due to the changes in the anthropogenic emissions we have made a budget for Europe and China for 1985 and 1996 (Table 10). In China the dry deposition and O3 oxidation percentages and the emissions increase while they decrease in Europe. The H2O2 oxidation percentage shows the opposite pattern, it decreases in China and increases in Europe. The amount of SO2 not oxidized to sulfate, i.e., wet and dry deposition, increases in China, while it decreases in Europe. In other words, when the emissions increase over China, the fraction that is oxidized decreases, and hence the effect of the changes in the emissions upon formation of sulfate is damped. In the opposite case, when the emissions decrease over Europe the effect on sulfate is damped as well.

Table 10. Relative Importance (%) of the SO2 Loss Processes for One Region Located in Europe and One in China Results Using the 1985 and 1996 Anthropogenic Emission Inventory
ProcessChinaEurope
1985199619851996
Wet deposition, %1.51.72.22.1
Dry deposition, %53.355.663.160.8
Gas phase oxidation, %8.38.56.36.2
H2O2 aqueous phase oxidation, %30.527.013.116.5
O3 aqueous phase oxidation, %3.84.47.36.8
HO2NO2 aqueous phase oxidation, %1.31.42.02.4
Catalytic, %1.21.36.05.2
Stratosphere, %0.10.1--
Total, %100.0100.0100.0100.0

[52] Two factors influence the sulfur cycle when we study the 1985 case versus the 1996 case; (1) changes in the emissions of sulfur and (2) changes in oxidants due to changes in emissions of oxidants and oxidant precursors (NOx and hydrocarbons). We performed model tests where we first changed the emissions of sulfur only, then changed the emissions of NOx and hydrocarbons only and finally changed all emissions (i.e., 1996 inventory). When we changed the emissions of NOx and hydrocarbons only, O3 increased by up to approximately 14 ppb in parts of China (in July, 955 hPa layer). On the other hand, an increase of the emissions of sulfur in China will imply that the oxidation of SO2 by O3 will get more important. Hence increased sulfur damp the effect on O3 of increased NOx and hydrocarbons. When we changed all anthropogenic emissions O3 increased by up 10 ppb in China in July. The sulfur reactions had therefore an impact on O3 corresponding to a decrease in O3 of 4 ppb.

[53] We apply a fixed pH = 4.5 in our model (see section 2.3). A sensitivity test has been performed with a fixed pH of 4. The largest effect of reduced pH is found in NH winter in layers 3–5 (200–1200 m), for Europe this typically leads to a 2–5% increase of SO2 (slower loss) and 5–10% decrease of sulfate in January in these layers, the values at the ground are hardly affected (no clouds in lowermost layers). The largest effect in absolute numbers is found in China (1200 ppt increase in SO2 and 700 ppt decrease in sulfate). Calculating pH in this region is not a straightforward task though. Observations of pH from two stations in China (L. Yu, personal communication, 2003) show pH between 4 and 6.5 for 1996, indicating an important role of alkali mineral dust. If pH is lower than 4.5, then the rate kO3 (Table 5) would be lower than in the model. The damping of 4 ppb on O3 from SO2 would then be an upper limit, since in the real atmosphere O3 would oxidize SO2 more slowly. The atmospheric mineral dust in China will counteract increased acidity due to enhanced sulfate, but models today lack all the compounds necessary to calculate correct pH over China.

4.1.3. Changes in OH Oxidation

[54] Gas phase oxidation by OH will lead to formation of new particles whereas aqueous phase oxidation and catalytic oxidation by metals take place on existing particles. The changes in sulfate particle formation could have implications for the formation of clouds and for its radiative properties. Figure 20 shows the total column of sulfate for July 1996 originating from oxidation by OH (compare with Figure 9 that shows column of total sulfate) and relative change from 1985 to 1996. Half of the sulfate over the Arabic Peninsula is a result of oxidation by OH. Sulfate from OH oxidation has increased with over 90% in Southeast Asia from 1985 to 1996, the decrease in Europe is about 60%. We must stress that the 90% increase in Southeast Asia is an upper limit since some of this new sulfate may stick to already existing particles. Two effects contribute to increased importance of OH oxidation: (1) a southward shift of the sulfur emissions toward regions with more incoming solar radiation and hence more OH (2) changes in the emission of oxidants and oxidant precursors. Globally more SO2 is oxidized by OH in the 1996 run than in the GEIA 1985 run. Sensitivity tests changing the emissions of sulfur only and then the emissions of NOx and hydrocarbons only show that changing the emissions of sulfur has the greatest impact on the fraction of SO2 lost by OH (see Table 11). Changing the emissions of NOx and hydrocarbons only do increase the OH oxidation, but not to the same extent. We must stress that there are secondary nonlinear effects included in these numbers, but we can however conclude that the increased oxidation by OH is due to a southward shift of the emissions of sulfur.

Figure 20.

(top) Monthly average total column of sulfate for July 1996 that originates from oxidation by OH, unit: μg(SO42−) m−2 and (bottom) percent change in sulfate from OH oxidation from 1985 to 1996 (100 × [1996 ÷ 1985]/1985).

Table 11. Total Loss of SO2 by OH Gas Phase Oxidation (Tg(S)) and Percentage Loss of SO2 by OH Gas Phase Oxidation (%)
Model RunTotal Loss of SO2 by OHPercentage Loss of SO2 by OH
1996 standard run7.918.74
GEIA 19857.478.32
Change sulfur emissions only7.938.76
Change NOx and hydrocarbons emissions only7.568.42

4.2. Comparison With Other Studies

[55] Total sources and sinks, burden and lifetime of SO2 and sulfate for the OsloCTM2 model and four other models are listed in Table 12. These models represent different sulfur schemes; Chin et al. [1996] and Restad et al. [1998] use prescribed oxidants, Koch et al. [1999] semiprognostic H2O2, and Roelofs et al. [1998] use a fully coupled scheme. We see that the OsloCTM2 gives smaller burden and shorter lifetime than the other models. One obvious reason for this is that our dry deposition is very effective (46% loss of SO2). Effective dry deposition will also make less SO2 available for oxidation. Compared to zonal averages from these four other studies our model have more DMS and SO2 confined to the boundary layer, i.e., our vertical transport is slower. Comparing sulfate, which is largely a secondary component, our results agree better by showing somewhat enhanced values with height in NH midlatitudes. Chin et al. [1996] report 21% SO2 removal by wet deposition. Owing to low solubility only 2% of SO2 is lost by wet deposition in our model, in agreement with Koch et al. [1999] and Roelofs et al. [1998]. Chin et al. [1996] let SO2 react with H2O2 in the aqueous phase and then remove it and count this as wet deposition of SO2, whereas we would count this as oxidation by H2O2. Our global loss by H2O2 is 30%, a percentage well below all these four other studies. The low fraction reported by Koch et al. [1999] (39%) is probably due to an oxidation limitation, while the high fraction by Roelofs et al. [1998] (48%) is partly due to their low fraction lost by dry deposition (16%), hence more SO2 is available for oxidation. There are two possible explanations for our low fraction of H2O2 oxidation: (1) we have included other pathways (catalytic, O3, HO2NO2) and (2) other models use prescribed fields of H2O2 and replenish H2O2 every time step while our model experiences an oxidation limitation and deplete H2O2. It will take a significant time before the levels of H2O2 are restored depending on the levels of HO2 (reaction R4, Table 3). The oxidation limitation is also seen if we study the SO2/sulfate ratio in industrialized areas in NH. In January the SO2/sulfate ratio exceeds 10 in Europe, in July it is between 1 and 5.

Table 12. Total Sources and Sinks and Burden and Lifetime for SO2 and Sulfate Using the GEIA 1985 Emission Inventory Comparison With Four Other Model Studies
 This WorkChin et al. [1996]Restad et al. [1998]Koch et al. [1999]Roelofs et al. [1998]
SO2
Total sources, Tg(S)/year89.295.691.980.490
Anthropogenic emissions67.165.165.264.663.7
Biomass burning2.32.32.02.32.3
Volcanoes8.06.78.73.57.8
Oxidation11.821.516.010.016.2
Total sinks, Tg(S)/year89.295.691.780.490.0
Dry deposition41.326.633.035.516.0
Wet deposition1.519.98.20.2(<1)
Gas phase oxidation (OH)7.57.511.013.116.2
Aqueous phase oxidation38.941.639.731.657.8
SO2 burden/lifetime0.26 Tg(S)/1.06 days0.34 Tg(S)/1.3 days0.42 Tg(S)/2.0 days0.56 Tg(S)/2.6 days0.61 Tg(S)/2.4 days
 
Sulfate
Total sources, Tg(S)/year49.949.150.746.678.0
Anthropogenic emissions3.5--1.93.3
SO2 oxidation46.449.150.744.774.7
Total sinks, Tg(S)/year49.949.150.746.678.0
Dry deposition7.45.68.69.217.0
Wet deposition42.543.542.137.461.0
Sulfate burden/lifetime0.50 Tg(S)/3.69 days0.53 Tg(S)/3.9 days0.62 Tg(S)/4.5 days0.73 Tg(S)/5.7 days0.96 Tg(S)/4.7 days

[56] We get a relatively short global lifetime of sulfate in the OsloCTM2 compared to other models (3.69 days). The levels of sulfate are very dependent upon the wet removal. The wet removal processes in the OsloCTM2 are represented in a very realistic way with 3-D cloud data and 3-D precipitation from the IFS model. These meteorological data compare well with observations [Gregory et al., 2000]. Hence our model calculated lifetime should represent the atmospheric lifetime of sulfate more realistic than models with a simpler parameterization of the wet removal.

[57] Like Roelofs et al. [1998] we have made a test using prescribed oxidants from a run without sulfur and compared it with the results using coupled oxidants. Our results agree well with their findings. Using prescribed oxidants affects primarily the results in NH winter, when there is strong oxidation limitation. By using off-line oxidants H2O2 oxidation is most affected and becomes more important, both in relative importance (%) and total loss, O3 oxidation and dry deposition become less important while OH oxidation is little affected (slightly less important). Our results for January show that H2O2 oxidation efficiency increase from 2.18 Tg(S) (online version, 25.6% of total loss of SO2) to 3.46 Tg(S) loss (40.7% of total loss of SO2) using prescribed oxidants. Global January mass of SO2 decrease from 0.34 Tg(S) to 0.23 Tg(S) while sulfate increase from 0.55 to 0.57 Tg(S). This test shows that the oxidation limitation is strongest for H2O2.

5. Conclusion

[58] In this study a sulfur cycle chemistry scheme with five components is included in the OsloCTM2 with all the relevant emissions, oxidation and deposition processes with particular emphasis on DMS, SO2 and sulfate. An improvement compared to previous studies is that the aqueous phase oxidation of SO2 to sulfate by O3, H2O2 and HO2NO2 is calculated interactively with the oxidants. Therefore changes in the emissions of oxidant precursors lead to an immediate response in oxidants compounds. Likewise we can study how the changes in sulfur affect the oxidants (small although except in some local spots). Comparisons with observations in remote oceanic regions and in anthropogenic regions used for model validation reproduce the observations mostly within a factor 2 although some discrepancies exist. The model overestimates SO2 and underestimates sulfate in Northern Hemisphere winter, due to missing oxidation pathways in wintertime, especially by H2O2. The model seems to give good agreement with observations in Europe (EMEP network), while SO2 is overestimated in North America (IMPROVE network) and somewhat underestimated in background areas (PEM-Tropics A). A global budget based on interactive chemistry is presented. Nearly half of SO2 is lost by dry deposition. For sulfate wet deposition is responsible for 85% of the loss. Oxidation of SO2 to sulfate occurs mainly via H2O2 (32% of total loss of SO2), followed by OH (9% of total loss) and O3 (7%). Regional and seasonal oxidation is somewhat different, H2O2 is the most important loss pathway for SO2 in background oceanic areas, whereas dry deposition is most important in anthropogenic areas. The anthropogenic emissions of SO2 have changed substantially from 1985 to 1996, with global emissions remaining virtually unchanged, while geographical patterns change significantly. The emissions have decreased in Europe and former USSR, due to cleansing technology and economic recession respectively, and have increased in Asia due to economic growth.

[59] For 1985 we estimate a global lifetime of 1.06 days for SO2 and 3.7 days for sulfate compared to 0.99 days and 3. days in 1996. The global burdens of SO2 and sulfate are 0.26 and 0.50 Tg(S) in 1985 and 0.25 and 0.53 Tg(S) in 1996, respectively. We find that in China with increasing emissions, the fraction that is oxidized to sulfate decreases and dry deposition increases, hence the rate of conversion to sulfate is damped. The situation is reversed in Europe where reduced emissions cause the fraction converted to sulfate to increase. The levels of oxidants are very important for the sulfur cycle, but the sulfur cycle does only have a small influence on the oxidants except in limited regions, notably Southeast Asia. In China where increased emissions of NOx and hydrocarbons enhance O3, increased emissions of SO2 inhibit O3 increase. The SO2 effect account for a 4 ppb reduction of O3. Finally the southward shift in the emissions from 1985 to 1996 increases the importance of OH as an oxidizing agent, particularly in Southeast Asia. This will affect the formation of new sulfate particles and hence affect radiative properties in this region.

Appendix A:: Description of the Convective Wash-Out

[60] At a level L, if we have entrainment, we calculate first the mass of air in the elevator (unit: kg):

equation image

Likewise we calculate mass of water at level L:

equation image

Q(L) is specific humidity at level L (unit: kg/kg).

[61] Then we find partial pressure of water vapor, before condensation:

equation image

Here p(L) is pressure at level L, ɛ = Rd/Rv = 0.622.

[62] Need temperature of the elevator, find first potential T of the environment:

equation image

Use mass weighted potential T to find potential T of the elevator:

equation image

Then we find temperature of the elevator:

equation image

Find then saturation partial pressure of water, es:

equation image

[Rogers and Yau, 1989, formula 2.17]. Tx = Televator(L) − 273.15, i.e., T of the elevator in degrees C. If etot > es then we have condensation and we do an iteration procedure to calculate the true temperature of the elevator, including the release of latent heat.

[63] First estimate of reduction of water vapor pressure due to condensation:

equation image

Not all is going to condense due to release of latent heat, try 20% as a first approach

equation image

Then the increase in Televator due to latent heat release is

equation image

Lw is latent heat of vaporization at 0°C, 2.5 × 106J kg−1, cp is specific heat at constant pressure, 1004 J deg−1 kg−1.

[64] If ∣δT∣ > 0.01 K we repeat the procedure (similar to A7):

equation image
equation image

T′ is in degrees C. Remaining water in the gas phase after condensation is

equation image

Check if we still have supersaturation, δe is assigned a new value

equation image

If δe > 0, not all is going to condense, use 20% as a second approach (similar to equations (A9) and (A10)):

equation image
equation image

Here δT is updated:

equation image

We repeat the steps (A11)(A17) until ∣δ′T∣ < 0.01. Then we update temperature and potential temperature of the elevator:

equation image
equation image

Now we have found the temperature of the elevator with a mixture of water vapor and droplets. We find new Tx and es using the updated Televator(L) in A7 and then find new δe using A8. The liquid water content (LWC) in the elevator in level L is consequently

equation image

Rain(L) is convective rain released from level L, given as an input parameter. We now know the fraction of water that is removed by convective precipitation. The amount of soluable tracer removed by wet deposition can now be calculated. A similar procedure is performed in case of condensation during detrainment.

Acknowledgments

[65] This work has received support from DOE through the project “Ozone as a climate gas” and from the Norwegian Research Council through the project RegClim and through a grant of computing time. The authors acknowledge the Canadian National Atmospheric Chemistry (NAtChem) Database (Julie Narayan and Bill Sukloff) and its data contributing agencies/organizations (IMPROVE network) for the provision of the data for year 1996 used in this publication. Finally, the comments and suggestions from two anonymous reviewers improved the manuscript substantially.

Ancillary