Journal of Geophysical Research: Atmospheres

A scheme for process-tagged SO4 and BC aerosols in NCAR CCM3: Validation and sensitivity to cloud processes

Authors


Abstract

[1] A life cycle scheme for sulfate (SO4) and black carbon (BC) is implemented in an extended version of the National Center for Atmospheric Research (NCAR) Community Climate Model 3 (CCM3). The scheme includes emissions of dimethyl sulfide (DMS), SO2, and sulfate of natural and anthropogenic origins and emissions of BC from biomass burning and fossil fuel combustion. Chemistry and aerosol physics are parameterized based on prescribed oxidant levels and background aerosols of marine, continental, and polar origins. Aqueous chemistry depends on estimated exchange rate of cloudy and clear air. Particulate SO4 and BC are tagged by-production mechanisms for off-line reconstruction of aerosol optical and water activity properties. With emissions from International Panel on Climate Change (IPCC), calculations without feedback produce atmospheric turnover times (days) of 1.5 (SO2), 3.5 (SO4), and 4.7 (BC) for the year 2000 and 1.6 (SO2), 4.0 (SO4), and 4.7 (BC) for the year 2100 A2 emission scenario. The modeled SOx compounds compare within a factor 2 with observations at ground level in North America and Europe and for SO4 in the free troposphere. For BC, the ground-level concentrations are well within a factor 10 from observations over several regions. BC and SO4 are a factor 10 too low in Arctic winter, which can partly be linked to spurious low-level winter cloudiness. SO4 and BC are a factor 10 too high at ground-level low latitudes, and upper tropospheric SO2 is largely missing. These major model biases are caused by neglected transport and low scavenging efficiency in cumulus clouds. Cloud processes are discussed by sensitivity tests. SO4 and BC are found very sensitive to the vertical transport and scavenging in convective clouds. More research should aim at improved cloud parameterization schemes that address key processes associated with aerosols to reduce uncertainties associated with climate effects of anthropogenic aerosols.

1. Introduction

[2] Aerosols exert a direct radiative forcing on the climate system by scattering and absorbing solar radiation [e.g., Clarke et al., 1984; Valero et al., 1984; Charlson et al., 1991; Kiehl and Briegleb, 1993]. By acting as cloud condensation nuclei (CCN) particles also influence the optical properties of clouds and the release of precipitation, and thus exert an indirect radiative forcing [e.g., Twomey, 1977; Charlson et al., 1987; Albrecht, 1989; Wigley, 1989]. Natural tropospheric aerosols are abundant as primary marine (sea salt) and continental (mineral) submicron particles, and as secondary particles formed by transformation of biogenic and volcanic gases to particles. Anthropogenic aerosols consist of sulfate (SO4), nitrate, and carbonaceous matter, and are products from biomass burning and fossil fuel combustion and secondary photochemical products. Changed wind climate and land use may also change the airborne sea salt and mineral aerosols.

[3] There is increasing evidence that anthropogenic aerosols significantly disturb the forcing of the climate system, but quantified estimates are considerably more uncertain than for greenhouse gases [e.g., Penner et al., 1994; IPCC, 2001; Shine and Forster, 1999]. A particle's power to stay airborne, to attenuate solar radiation, and to become a CCN depends on its size, shape, internal structure, and composition. Particle diameters vary with at least two orders of magnitude, and in an air parcel each size class spans a wide range of composition and structure. Measurements of adequate detail will hardly ever become available and calculations from “first principles” are too computer demanding in climate models. Shortcuts involving parameter tuning cannot be avoided, which implies that the quality of nonmeasurable but calculated parameters remain uncertain. In this paper we also show that parameterized cloud dynamics contribute considerably to the uncertainty.

[4] Carbonaceous particles include light-absorbing black carbon (BC) and a huge class of organic carbon species (OC) with differing optical and solubility properties [e.g., Seinfeld and Pandis, 1998]. The TARFOX campaign regions [Hegg et al., 1997; Novakov et al., 1997] showed that OC can dominate the aerosol mass and optical depth relative to SO4. SO2 and SO4 (and later nitrate) have been measured and modeled in connection with acid rain studies for three decades, and annual anthropogenic emissions and boundary layer concentrations are regionally well known (for Europe, see www.emep.int). Carbonaceous aerosols have not received similar attention. The sources, speciation and physical properties of primary and secondary OC are barely known. Efforts have been made for emissions of BC and bulk OC [Penner et al., 1993; Cooke and Wilson, 1996; Liousse et al., 1996; Cooke et al., 1999] and for OC properties [e.g., Saxena and Hildeman, 1996; Dick et al., 2000].

[5] The short residence time of particles (several days to a week) hampers well-mixed distributions in the troposphere, but transport in air is important for the spatial distribution. In order to map this distribution, ground-level SO4 observation data are available in Europe and North America, but are too sparse remotely and in the free troposphere. BC and OC measurements are even sparser. Since measurements are inadequate model calculation is necessary. At least a dozen global models for sulfur are available [Langner and Rodhe, 1991; Pham et al., 1995; Dastoor and Pudykiewicz, 1996; Chin et al., 1996, 2000; Feichter et al., 1996; Chuang et al., 1997; Kasibhatla et al., 1997; Roelofs et al., 1998; Restad et al., 1998; Koch et al., 1999; Barth et al., 2000]. Global models for BC are fewer [Penner et al., 1993; Cooke and Wilson, 1996; Liousse et al., 1996; Cooke et al., 1999], the two latter including OC. In the work of Penner et al. [1998], Lohmann et al. [1999], and Koch [2001], BC, OC, and SO4 were all calculated, but only the two latter cases with physical interactions between them.

[6] The sulfur model intercomparison of Rasch et al. [2000b] found that models differed greatly in simulation of soluble species but that lack of upper-level measurements hampered firm quality assessments. The COSAM study [Barrie et al., 2001] evaluated 10 sulfur models, 3 of which were GCMs. Global models overpredicted seasonal surface SO2 by a factor of 2 or more, while seasonal surface SO4 were mostly within 20% of measurements. In winter there was a tendency to underpredict close to sources and overpredict remotely, which revives the idea of a missing oxidation mechanism of SO2 [Kasibhatla et al., 1997]. Too efficient vertical mixing from the boundary layer to the free troposphere were modeled in the GCMs, and this was considered a major source of uncertainty for the SO4 distribution in climate models.

[7] Often models with “bulk” formulations of oxidation and removal tuned for regional sulfur transport up to a few 1000 km, give better agreement with regional measurements than global models [e.g., Eliassen and Saltbones, 1983; Iversen, 1993; Christensen, 1997]. For GCMs deviations of meteorological fields from reality and coarser grid resolution are possible reasons, but the same features result from CTMs that use observed meteorology and higher resolution. Hence it is probable that more explicit and physically based process formulations uncover gaps of knowledge that are hidden in tuned “bulk” parameters. An enlightening example was shown in COSAM, where the hemispheric model of Christensen [1997] compared far better with observations than the 9 process-resolving models.

[8] “Bulk” formulations of aerosol microphysics are inadequate in order to estimate optical properties and CCN activation realistically. Even for a pure SO4 aerosol, Kiehl and Briegleb [1993], who assumed size distributions and spectral optical Mie parameters, obtained a factor 2 lower direct forcing than obtained by Charlson et al. [1991] with “bulk” aerosol optical properties. Likewise, SO4 contributions to CCN depend on the fraction of SO4 produced in gas phase. Jacobson [2001] demonstrated the importance of physical mixing between aerosol species. Koch [2001] determined the mixing state of BC by condensation of gas-phase SO4, but physical interaction mechanisms between aerosol species are largely missing in present climate models.

[9] We use a version of the National Center for Atmospheric Research (NCAR) Community Climate Model 3 (CCM3) [Kiehl et al., 1998] extended with the cloud treatment of Rasch and Kristjánsson [1998]. Except for the treatment of aerosols we use the same model code as used by Barth et al. [2000], Rasch et al. [2000a], and Kiehl et al. [2000]. A life cycle scheme for SO4 with precursors and BC is implemented. Tagging of concentrations according to production mechanisms enables reconstruction of particle size distributions and compositions. Internal mixing between SO4 and BC is entirely caused by coagulation, for which prescribed background primary particles of marine, continental or polar types determine the efficiency. The sulfur chemistry uses prescribed oxidants, with a minimum oxidation rate assumed in clouds [Kasibhatla et al., 1997]. Slightly simpler versions of the scheme were tested in a hemispheric-scale CTM by Seland and Iversen [1999] and Seland [2001]. Omitting OC and nitrate probably causes underestimated CCN [e.g., Rivera-Carpio et al., 1996]. BC is included due to its absorptivity of light. Pure BC is hydrophobic but in internal mixtures the absorptivity increases and CCN production may be influenced [Ackerman and Toon, 1981; Haywood and Ramaswamy, 1998; Myhre et al., 1998; Jacobson, 2001]. It is thus crucial that SO4 and BC are modeled together.

[10] Model validation includes checking the agreement with calculations with observations, and comparing budgets and distributions with those of other models. For sulfur we in particular compare with the models of Barth et al. [2000], Rasch et al. [2000a], Kasibhatla et al. [1997], Koch et al. [1999], and Chin et al. [1996, 2000]. For BC our main references are the models of Liousse et al. [1996], Cooke et al. [1999], and Koch [2001].

[11] Several sensitivity tests for cloud-related processes involving transport, chemistry and scavenging are run. Inspired by COSAM and biases presented in Barth et al. [2000] and Rasch et al. [2000a], we consider in particular the vertical mixing and scavenging in deep cumulus clouds. Finally, two emission scenarios are used, and intercontinental transport of SO4 is briefly discussed.

[12] This paper presents no feedback from modeled SO4 and BC onto the GCM physics and dynamics. Radiative forcing is estimated by Kirkevåg and Iversen [2002] (direct) and by Kristjansson [2002] (indirect). Preliminary results were given by Iversen et al. [2001].

2. The Model

2.1. The GCM

[13] The model is NCAR CCM3.2 [Kiehl et al., 1998] extended with the cloud scheme of Rasch and Kristjánsson [1998]. This is only a brief summary since no changes are introduced. The resolution is T42 with 18 levels and a hybrid (η) vertical coordinate with upper lid at 2.9 hPa. Dynamic time step length is 20 min, and radiation flux densities are updated hourly. Semi-Lagrangian advection is used for positive definite scalars with extra control for global conservation. The model parameterizes subgrid boundary layer transport by local gradient diffusivity under stable and near-neutral conditions and a nonlocal scheme under convective conditions. For shallow and midlevel convective clouds the scheme of Hack [1994] is used.

[14] For deep moist convection the model uses the plume-ensemble mass flux scheme with “convective available potential energy (CAPE) closure” of Zhang and McFarlane [1995]. Updraft mass fluxes are determined by a presumed exponential destruction rate of the CAPE in a grid column. Detrainment of updraft cloudy air only takes place in thin layers where plumes are neutrally buoyant. Downdraft mass fluxes are driven by precipitation production in updraft plumes, start below the updraft detrainment layers and detrain below the cloud.

[15] The long-wave radiation code explicitly includes clouds and water vapor. Concentrations of greenhouse gases are prescribed. For solar radiation the δ-Eddington scheme is used. For this model version without feedback from the SO4 and BC on model physics, standard radiative properties of clouds and aerosols are prescribed. Cloud fields follow the parameterization of Rasch and Kristjánsson [1998], and a passive boundary layer aerosol with optical properties by Kiehl and Briegleb [1993] is prescribed.

[16] The prognostic cloud scheme of Rasch and Kristjánsson [1998] estimates precipitation autoconversion rates and cloud radiation interactions from assumed droplet number concentrations and effective radius. This feature makes the model particularly well suited for linking an on-line calculated aerosol to indirect climate effects in a next step.

[17] Transport in convective clouds. The results of Barth et al. [2000] underestimated ground-level SO4 typically a factor 2 at remote sites and 2–5 in source regions. Ground-level SO2 in source regions typically agreed well within a factor 2 with a tendency to overestimate in North America, while no general bias was seen for sulfur wet deposition. For the sparse measurements in the Pacific free troposphere, SO4 was overestimated up to a factor 5 in the upper part and a factor 1.5–2 in the middle part, while there was no general bias for SO2. A missing oxidation mechanism in winter could explain parts of the SO4 errors in source regions, but Barth et al. [2000] stated that vertical transport out of the boundary layer had to be a part of the explanation. Too efficient vertical transport of boundary layer air was also pointed out as a major source of error in GCMs from COSAM [Barrie et al., 2001].

[18] This vertical transport can be due to convective clouds whose vertical mass fluxes involve major portions of boundary layer air over a model time step. If detrainment of air entirely takes place where updraft plumes become neutrally buoyant [Kiehl et al., 1998], boundary layer contaminants are efficiently transferred to the upper troposphere. Koch et al. [1999] found that over the polluted regions in the Northern Hemisphere, moist convection in their model delivered more SO2 to the free troposphere than was oxidized. In the upper troposphere SO4 is to a large extent produced in gas phase, and is less susceptible to scavenging than in the low-level SO4. Koch et al. [1999] had similar biases as Barth et al. [2000], but calculated more SO2 and SO4 throughout the Northern Hemisphere due to slower depletion (see Table 5).

[19] To contrast with Barth et al. [2000] and Rasch et al. [2000a] we simply neglect vertical transport of dimethyl sulfide (DMS), SO2, SO4, and BC in convective clouds. Hence one should expect overestimated SO4 (and BC) at low levels near source regions but smaller than the free troposphere overestimates by Barth et al. [2000] since scavenging is more efficient in lower layers. It is beyond the scope of this paper to develop a parameterization scheme better suited for contaminant transport but does not destroy the other beneficial properties of the scheme [Kiehl et al., 1998]. Instead we investigate the significance of the convective transport in more depth in section 4.1.

2.2. Emission Data

[20] We use sulfur emission scenarios compiled for the IPCC [2001] for 2000 and 2100 (International Panel on Climate Change (IPCC) SRES A2) [Nakicenovic et al., 2000]. For BC we use data based on the works of Liousse et al. [1996] and Penner et al. [1993] for 2000, while data for 2100 are scaled from CO projections [IPCC, 2001, chapter 5]. These data were also used by Koch [2001]. The global emissions of natural sulfur (DMS: 25.3 Tg(S)/a; volcanic SOx: 4.8 Tg(S)/a) are assumed unchanged over the century. Anthropogenic SOx are assumed to 69.0 Tg(S)/a for 2000 and 60.3 Tg(S)/a for 2100. The original emission data for BC (12.4 Tg(C)/a in 2000; 28.8 Tg(C)/a in 2100) do not separate emissions from biomass burning and fossil fuel combustion. We have crudely estimated this using the same relative separation on a monthly basis as Cooke and Wilson [1996].

[21] Seasonal variations are given except for anthropogenic sulfur. Traditionally a seasonal variation due to domestic heating is assumed in Europe, but considerable structural changes since 1990 has caused that there are no overall reliable data available, and certainly not for 2100. Hence we stay with the annual IPCC data even though some systematic errors may occur.

[22] The emission heights above ground are not explicitly known. DMS is injected at ground level, while SOx emissions are distributed in the three lowest model levels (below ca. 750 m) with weights 0.25, 0.5, and 0.25. Since volcanic emissions exclude eruptions, they are injected in the upper part of the boundary layer between ca. 350 and 750 m (third layer above ground). BC emissions from fossil fuel stem to a larger extent than SOx from traffic, domestic heating and small-scale industry, and are injected in the two lowest levels (below ca. 350 m) with weights 0.75 (in the lowermost) and 0.25. BC emissions from biomass burning are assumed to be ground level. It would probably be more correct to distribute these emissions higher due to rising fire plumes [Liousse et al., 1996].

2.3. Aerosol Physics

[23] The scheme for the airborne concentration and deposition of the aerosol species is an adjusted and slightly extended version of that presented by Seland and Iversen [1999]. SO4 and BC are tagged according to chemical and physical production. Figure 1 shows a schematic of the 9 calculated compounds, and Table 1 gives all transformation and deposition rates with references. In the following we use acronyms defined there.

Figure 1.

A schematic for the sulfur and BC scheme implemented in CCM3. Q(DMS), Q(SO2), Q(SO4), Q(C), Q(ff), and Q(bb) are emissions of DMS, SO2, SO4, BC, fossil fuel BC, and biomass burning BC, respectively. Calculated concentrations that are historical variables in the model are given in rectangles. MSA is an immediate loss of DMS and are not stored in the model. SO4(n) and C(n) are externally mixed SO4 and BC in nucleation/Aitken mode. Cx(a) is externally mixed accumulation mode BC. SO4(a1), SO4(a2), and SO4(a3) are internally mixed SO4 produced in different ways as illustrated and explained in Table 1. C(a) is internally mixed BC. All components, except DMS, are subject to dry and wet deposition. “Gas phase” signifies SO2 oxidation in clear air and “wet phase” in cloud droplets. New SO4 particle formation is produced in gas phase with f = 0.05. Coag(dry,n) is clear-air coagulation of nucleation/Aitken mode particles. Coag(cl,n) and Coag(cl,a) are coagulation between cloud droplets and nucleation/Aitken mode particles and accumulation-mode particles, respectively.

Table 1. Transformation and Deposition Rates in the Model
Clear-air coagulation:
SO4(n) → SO4(a2)Coag(dry,n)rates by Seland and Iversen [1999, Table 4] 
C(n) → C(a)Coag(dry,n)rates by Seland and Iversen [1999, Table 4] 
 
Cloudy-air coagulation:
SO4(n) → SO4(a3) Coag(cl,n) = 1 h−1 = 2.78 × 10−4 s−1rate estimate by Ogren and Charlson [1983] 
C(n) → C(a)Coag(cl,n) = 1 h−1 = 2.78 × 10−4 s−1rate estimate by Ogren and Charlson [1983] 
SO4(a1) → SO4(a3)Coag(cl,a) = 0.1 h−1 = 2.78 × 10−5 s−1rate estimate by Ogren and Charlson [1983] 
SO4(a2) → SO4(a3)Coag(cl,a) = 0.1 h−1 = 2.78 × 10−5 s−1rate estimate by Ogren and Charlson [1983] 
Cx(a) → C(a)Coag(cl,a) = 0.1 h−1 = 2.78 × 10−5 s−1rate estimate by Ogren and Charlson [1983] 
 
Clear-air chemistry:
SO2 + OH → f SO4(n) + (1 − f) SO4(a1), f = 0.05,rate by Barth et al. [2000] 
DMS + OH → SO2kDMS = α(1.2 × 10−11 cm3 mol−1 s−1)e260 K/T; α = 0.66Seinfeld and Pandis [1998] and Yin et al. [1990] 
DMS + NO3 → SO2swift in darkness in the ABL north of 10°Nsee main text 
 
Cloudy-air chemistry:
SO2 → SO4(a3) keff = keff (kwet) = effective rate over model time step (Δt)as in Seland and Iversen [1999] 
where: kwet = kmin + kO3 + kH202assumed min. rate 
  kmin = 5.6 × 10−6 s−1 
  kO3 = Vw HO3 H* (kO0aSO2.H2O + kO1aHSO3 + kO2aSO2) [O3] 
  kH2O2 = 0.8HH2O2 (1 + 0.8HH2O2)−1 [H+] (1 + K[H+])−1 Vw H* kH1 aHSO3 [H2O2] 
  kO0 = 2.4 × 104 M−1 s−1as in Seland and Iversen [1999] 
  kO1 = (3.7 × 105 M−1 s−1) exp[5530 K(1/298 K − 1/T)]Seinfeld and Pandis [1998, Table 6.A.7] 
  kO2 = (1.5 × 109 M−1 s−1) exp[5280 K(1/298 K − 1/T)]Seinfeld and Pandis [1998, Table 6.A.7] 
  kH1 = (7.5 × 107 M−1 s−1) exp[4430 K(1/298 K − 1/T)]Seinfeld and Pandis [1998, Table 6.A.7] 
  H* = HSO2R0TDs  
  R0 = 8.314 J mol−1 K−1  
  Ds = 1 + Ks1[H+]−1 + Ks1Ks2[H+]−2  
  HSO2, HO3, HH2O2 = Henry law constantsSeinfeld and Pandis [1998, Tables 6.2 and 6.3] 
  [H+] = 10−pH = proton concentration in cloud water, pH = 4.0  
  [O3], [H2O2] = air concentrations (partial pressure)Berntsen and Isaksen [1997] 
  aSO2.H2O = Ds−1  
  aHSO3 = Ks1[H+]−1Ds−1  
  aSO2 = Ks1Ks2[H+]−2Ds−1  
  Ks1 = (1.3 × 10−2 M atm−1) exp[1960 K(1/T − 1/298 K)]Seinfeld and Pandis [1998, Table 6.A.1] 
  Ks2 = (6.6 × 10−8 M) exp[1500 K(1/T − 1/298 K)]Seinfeld and Pandis [1998, Table 6.A.1] 
  Vw = liquid water volume mixing ratio  
  M = mol l−1  
 
Dry deposition:
All sulfur componentsLand use dependent, Resistance methodfully as in Barth et al. [2000] and Rasch et al. [2000a] 
BCSame as for SO4 particles  
 
Wet deposition:
qt(z)w = − [(cvEic ΔP(z) + caEbcP(z))(Vw Δz)−1]q;
where: qt(z)w = rate of depletion of concentration q at height z due to wet deposition
  Eic = in-cloud scavenging efficiency (values below);
  Ebc = below-cloud scavenging efficiency (values below);
  P(z) = precipitation rate (mm/h) from aloft in area fraction ca;
  ΔP(z) = precipitation rate produced in level z in volume fraction cv.
  Δz = layer thickness (m)
SO2Eic = keff τpEbc = 0τp = 1000 s, Seland and Iversen [1999]
SO4(n)Eic = 0Ebc = 0.2based on Hobbs [1993]
SO4(a1)Eic = 0.6Ebc = 0.2based on Hobbs [1993]
SO4(a2)Eic = 0.6Ebc = 0.2based on Hobbs [1993]
SO4(a3)Eic = 0.8Ebc = 0.1based on Hobbs [1993]
C(n)Eic = 0Ebc = 0.2based on Hobbs [1993]
Cx(a)Eic = 0Ebc = 0.2based on Hobbs [1993]
C(a)Eic = 0.5Ebc = 0.2based on Hobbs [1993]

[24] The purpose of tagging is to construct size distribution and composition a posteriori. Details of the reconstruction process are described by Kirkevåg and Iversen [2002]. The outset is presumed size distributions of dry primary particles of a marine, continental or polar type. These background aerosols are (so far) not transported, and each type is fixed for surface types (continents, oceans, ice cover) and altitude. They are modified by external and internal mixing with model-calculated SO4 and BC, and by humidity swelling according to Köhler theory and actual relative humidity (Rh). Externally mixed SO4 and BC have presumed dry size distributions modified by humidity swelling. In Figure 1 and Table 1 these compounds are SO4(n), C(n) (nucleation/Aitken-mode SO4 and BC), and Cx(a) (accumulation-mode BC). Internal mixing is caused by condensation of gaseous sulfuric acid (SO4(a1)), by clear air coagulation with externally mixed SO4 (SO4(a2)) and BC (C(a)), or by SO4 production in cloud droplets or coagulation with cloud droplets that are subsequently evaporated (SO4(a3) and C(a)). For BC we do not keep separate budgets for clear-air and cloudy-air coagulation. Self-coagulation is neglected except for the implicit assumptions behind the externally mixed SO4(n), C(n), and Cx(a).

[25] The assumed condensation fraction producing SO4(a1) is (1 − f) = 95%, which is the fraction of gaseous sulfuric acid that does not nucleate to SO4(n). Nucleation is in reality a function of temperature, Rh, and the available surface area of existing particles, and bursts may occur after aerosol scavenging under humid conditions [Raes et al., 1995; Zhang et al., 1998]. Our constant value of f = 5% reflects the problem of defining representative values in a climate model, but long-term SO4 budgets seems negligibly sensitive to the exact choice of f [Seland and Iversen, 1999]. Directly emitted SO4 is also ascribed to SO4(n).

[26] Clear-air coagulation rates (Coag(dry,n)) are precalculated assuming normalized accumulation-mode size distributions and number concentrations according to background aerosol type and height above ground [Jaenicke, 1993]. The transformation rates for nucleation/Aitken-mode particles are obtained by using the Fuchs formula for Brownian diffusion [Seinfeld and Pandis, 1998, p. 661], and are given by Seland and Iversen [1999, Table 4].

[27] Cloudy-air coagulation rates are larger than clear-air rates. There is coagulation between nucleation/Aitken-mode particles and cloud droplets (Coag(cl,n)) and between accumulation-mode particles and cloud droplets (Coag(cl,a)). The fraction of droplets that does not precipitate, leaves accumulation-mode particles behind after evaporation. Values are given by Ogren and Charlson [1983].

2.4. Sulfur Chemistry

[28] SO4 precursors are emissions of DMS and SO2. Only a minor fraction (∼2%) of the sulfur emissions is SO4 and the bulk part of atmospheric SO4 is oxidized precursor gases. The major oxidizing agents OH, H2O2, and O3 are prescribed from the global Oslo CTM I [Berntsen and Isaksen, 1997] as monthly averages with 8° × 10° resolution in 7 layers. The oxidants are calculated without sulfur chemistry in the model. OH is reduced to zero during darkness but increased in daytime to keep diurnal averages right. It is also reduced with by factor 3 in clouds during daytime as a consequence of reduced UV radiation. An important improvement compared to the work of Seland and Iversen [1999] is that all reactions here depend on temperature.

[29] For oxidants we simply assume the same levels in 2100 as in 2000 as we do not have reliable data for 2100. Also we use the same greenhouse gases and sea surface temperatures for 2100. Effects of changed temperature and oxidant levels on sulfur chemistry would be interesting if projections of oxidants were available. Hence our 2100 scenario is only a sensitivity test for different emission patterns, and in any case it is without feedback to the dynamics.

[30] Of the emitted DMS 34% is immediately removed as methane sulfonic acid (MSA) and never contributes to DMS, SO2, or SO4 in the model. The remaining fraction yields SO2 by oxidation with OH in daylight and with NO3 in darkness. The OH reaction is the H-abstraction pathway of Yin et al. [1990] [see also Seinfeld and Pandis, 1998, p. 316]. The nitrate radical reaction is assumed to only take place in darkness over continental areas north of 10°N with a reaction rate decreasing linearly from 1 h−1 on the ground to zero at 2 km. The quite low SO2 yield is supported by Yin et al. [1990] and by measurements in tropical Pacific [Seinfeld and Pandis, 1998].

[31] SO2 reacts with OH and produces gaseous sulfuric acid in clear air, which form new externally mixed nucleation/Aitken-mode particles of sulfuric acid and water (f = 5%) or condenses on preexisting accumulation-mode particles (95%). Reaction rates are the same as used by Barth et al. [2000].

[32] SO2 oxidation in aqueous phase by H2O2 and O3 is efficient when liquid cloud droplets are available. Normally the H2O2 reaction dominates in source regions for SO2 where acid conditions often prevail. A pH of 4.0 in cloud droplets is presumed, hence oxidation by O3 may be underestimated if base cations are abundant. Simple reaction tests give that pH need to be higher than 5 before this reaction becomes comparable to the H2O2 reaction. Henry law constants, equilibria, and reaction rates (Table 1) are given by Seinfeld and Pandis [1998]. The effective H2O2 concentrations are lowered with a factor 0.8 to account for swift depletion in clouds [e.g., Barth et al., 1989]. In addition to these resolved reactions we add a constant in-cloud rate of kmin = 2% h−1 = 5.6 × 10−6 s−1 inspired by Kasibhatla et al. [1997], who used 10−6 s−1 in winter and 2 × 10−6 s−1 in summer everywhere. Biases for SO2 and SO4 are found also in other process resolving models and was discussed in COSAM [Barrie et al., 2001]. Candidates for added oxidation mechanisms are: the effects of using a constant pH is mentioned; oxidation by O2 catalyzed by Mn and Fe has been neglected; bulk water cloud microphysics may reduce the effective oxidation rate; and H2O2 predicted by coarse-resolution CTMs may be too low in the lower troposphere [Lohmann et al., 2001]. As demonstrated in sensitivity tests for H2O2 [Barth et al., 2000; Koch et al., 1999; Roelofs et al., 1998], a model that underestimates wet-phase oxidation rates amplify errors in SO2 and SO4 due to convective transport. It is yet uncertain which of the two possible sources of error dominates.

[33] As stated by Barth et al. [2000], oxidation rate is forced to decrease linearly with temperature from its calculated value kwet at 0°C to 0 at −25°C. Seland and Iversen [1999] showed that sulfur in the Northern Hemisphere is considerably sensitive to how oxidation rates are reduced in ice clouds.

[34] The effective in-cloud oxidation rate, keff, for a grid volume and a model time step is smaller than kwet if H2O2 or SO2 are efficiently consumed from the reservoir of air involved in the cloud production. This keff depends on replenishment of SO2 and H2O2. Some studies [Barth et al., 2000; Koch et al., 1999; Roelofs et al., 1998] use chemical replenishment of H2O2 and show that prescribed concentrations that are replenished every time step produce more SO2. We advocate that kinematic replenishment of cloudy air partly will counteract this effect. Generalizing Rodhe and Grandell [1981] crucial parameters are the fraction of time Lagrangian air parcels in a grid volume are cloudy, and the residence time of air parcels in cloudy and clear air. At present such parameters are not available from cloud schemes in climate models. As detailed by Seland and Iversen [1999] we have tabulated keff as a function of kwet and SO2 by solving a diffusive depletion equation. We have used a Lagrangian cloudy time fraction of 0.3 and a diffusive depletion rate for cloudy air of 1 h−1. For certain ranges of kwet values Keff ≈ a kwet, where a ≈ 0.65 for SO2 = 0.1 μg m−3 and kwet = 2–20 × 10−5 s−1, and a ≈ 0.02 for SO2 = 14 μg m−3 and kwet = 2–20 × 10−3 s−1. If kwet would cause exhausted SO2 over an hour or shorter, keff asymptotically approaches 1 h−1 = 2.78 × 10−4 s−1, but if H2O2 is quickly exhausted, keff approaches kmin.

2.5. BC Transformations

[35] BC can transform from an externally mixed hydrophobic to a hydrophilic internally mixed aerosol component. In the model this takes place by coagulation only (section 2.2). BC is emitted as primary particles by incomplete combustion. Fossil fuel emissions (Q(ff)) are assumed emitted as 90% nucleation/Aitken-mode particles (C(n)) and 10% accumulation-mode conglomerates created by self-coagulation in the exhaust (Cx(a)) [Ström et al., 1992]. This assumption strongly affects the rate at which BC turns from hydrophobic to hydrophilic [Seland and Iversen, 1999].

[36] Half of the BC emission from biomass burning (Q(bb)) is assumed emitted as hydrophilic, accumulation-mode particles (C(a)). Even though OC is not calculated in the model, this mimics that BC coagulates with OC and SO4 in the fire plumes [Liousse et al., 1996]. Assuming that OC in this case is fully soluble, this BC fraction is considered hydrophilic.

2.6. Dry and Wet Removal

[37] DMS is neither dry nor wet deposited. For dry deposition of SO2 and SO4 we use the same code as used by Barth et al. [2000]. This is the resistance analogy of Wesely [1989]. Aerodynamic and sublayer resistances are determined from parameters in the surface boundary layer. BC particles in the boundary layer are deposited with the same efficiency as SO4 particles.

[38] Wet deposition is calculated in full interaction with the cloud and precipitation schemes in the model. This is different from the off-line CTM of Seland and Iversen [1999] where the nonprecipitating cloud fraction had to be estimated. We use scavenging efficiencies that reflect assumed sizes and the hygroscopicity of the particles. Values for in-cloud and below-cloud scavenging are different (Table 1).

[39] In-cloud scavenging of SO2 is brought about by uptake, dissociation, and oxidation in cloud droplets as they are converted to precipitation. The scavenging efficiency is the effective oxidation rate multiplied with an assumed SO2 exposure time for precipitation elements (τP) of 1000s. Below clouds we assume saturated SO2 solutions in raindrops associated with exhausted oxidant levels, and hence no below-cloud scavenging of SO2.

[40] In-cloud scavenging of particles is partly taken care of by coagulation in cloudy air, hence scavenging efficiencies for nucleation/Aitken-mode particles and externally mixed BC is zero. For accumulation-mode soluble particles scavenging efficiencies smaller than one reflects the situation that not all particles are activated CCNs. Larger scavenging efficiencies for convective precipitation could be argued for, but so far we use the same values. Chosen values are estimated from figures assembled by Hobbs [1993]. In the work of Barth et al. [2000] the in-cloud scavenging efficiency was 1 for all aerosols. Below-cloud particle scavenging efficiencies are estimated from the work of Hobbs [1993, Figure 12]. The SO4(a3) mode is closest to the scavenging gap at diameters slightly smaller than 1 μm and has half the value of other modes (0.1). Barth et al. [2000] used this value for all SO4. Seland and Iversen [1999] found that SO4 and BC budgets are sensitive to the formulation of below-cloud scavenging.

[41] We use the same scavenging efficiencies for convective and stratiform precipitation, which may underestimate scavenging at low levels by convective precipitation. Barth et al. [2000] used the column maximum cloud fraction for in-cloud scavenging, which causes efficient scavenging by convective precipitation in particular. The fraction of air in a boundary layer grid volume that is exposed to in-cloud convective precipitation over a model time step, should be proportional to updraft mass fluxes. In this paper we use the cloud volume fraction determined in each layer, since we have neglected vertical transport in convective clouds, which is linked to this issue. This is further discussed in section 4.1.

[42] Wet deposition rates are reduced for increasing ice contents in clouds with the same linear dependence with temperature as the aqueous oxidation of SO2.

3. Global Distribution of SO4 and BC

[43] Only anthropogenic emissions of sulfur and BC vary between the two scenarios (2000 and 2100). In each case 5 years were simulated with data from the last three used for statistics. A 2-year spin-up is sufficient for the troposphere but not completely for the stratosphere [Rasch et al., 2000a]. Stratospheric influence is not discussed here, and we present results for the latter 3-year period.

3.1. Concentrations and Column Burdens 2000

3.1.1. Burdens

[44] Figure 2 shows zonal averages for annually averaged concentrations for DMS, SO2, SO4, and BC. The sulfur components can be directly compared with the work of Barth et al. [2000, Figure 1]. In the lower troposphere we have a factor 2 more DMS than obtained by Barth et al. [2000], otherwise we lack a plume in the upper tropical troposphere due to our neglected convective transport, and we have a factor 5–10 higher burdens in the middle and high northern latitudes. The latter is at least partly caused by a factor ∼2.5 higher North Atlantic emissions. A presumed MSA yield of 34% of the DMS emission is removed and never contributes to DMS or SOx. Barth et al. [2000] also used a (unspecified) MSA yield fraction for DMS, and our effective emission is more than 8% higher globally. This is insufficient to explain our larger DMS burden.

Figure 2.

Zonal averages of volume mixing ratios of the annually averaged DMS, SO2, sulfate, (pmol (mol air)−1), and BC (ng(C) m−3).

[45] In our calculations both SO2 and SO4 are confined to the lower troposphere. Compared to Barth et al.'s [2000], our boundary layer concentrations of SO2 are slightly larger in the northern midlatitudes, a factor 3–5 smaller in the Arctic and the tropics, and a factor 2 larger in of the Southern Hemisphere. Even more striking is our 10–100 times smaller concentrations in major portions of the middle and upper free troposphere, a feature confirmed by comparing with those obtained by Koch et al. [1999] and Chin et al. [2000].

[46] Our lower tropospheric SO4 concentrations are a factor 2 larger in the Southern Hemisphere and in the northern midlatitude; similar or slightly smaller in the Arctic; and up to a factor 5 larger in the tropics. The dominating difference is that Barth et al. [2000] typically have from 5 to 20 times larger concentrations in the upper half of the troposphere. We have injected volcanic emissions entirely in the boundary layer assuming no eruptions. Parts of the upper-level deficiencies may be caused by this assumption, but our neglecting convective transport is a more likely explanation. It is interesting to compare our calculations with the calculations of the two papers of Chin et al. [1996, 2000] because of significant differences between them that may be ascribed to smaller convective scavenging in the former. Chin et al. [1996, 2000] used off-line meteorological data, but while the latter paper used standard convective fluxes from the meteorological model, Chin et al. [1996] estimated these from 5-day statistics of convective events [Prather et al., 1987]. Our zonal average of SO4 is more similar to that of Chin et al. [1996], even though their burdens were a factor 2–4 larger in the lower Arctic and the middle to upper troposphere and slightly lower in the tropical boundary layer. On the other hand, Chin et al. [2000] calculated zonal averaged SO4 closer to that calculated by Barth et al. [2000] and Koch et al. [1999] with a minimum in the tropical upper troposphere. Our SO4 has a sort of minimum, but clearly no maximum above.

[47] Compared to that obtained by Liousse et al. [1996] we get typically a factor 1.5–2 larger boundary layer BC concentration in the tropics and northern midlatitudes, a factor 2 smaller in the Arctic, and similar concentrations in the Southern Hemisphere. In the upper tropical troposphere our concentrations are smaller by a factor 10, and elsewhere by typically less than a factor 1.5. To our knowledge Liousse et al. [1996] used a CTM without explicit moist convection fluxes and scavenging. Koch [2001], who included this, calculated a more pronounced minimum in the upper tropical troposphere than we did. This minimum vanished in a test without convective scavenging. In the Northern Hemispheric middle and upper troposphere we calculate similar BC levels as calculated by Liousse et al. [1996] but only ∼20% of those by Koch [2001]. One reason could be the transfer to hydrophilic BC, but since this is faster in the work of Koch [2001, Table 6] the probable explanation also in this case is our (and Liousse et al.'s [1996]) neglected transport in convective clouds.

[48] Figure 3 shows integrated column burdens with same units and isolines as presented by Barth et al. [2000, Plate 1]. We have higher burden of DMS in general, and in the Arctic this is partly due to larger North Atlantic emissions. Our SO2 column is generally lower in the Northern Hemisphere and the Arctic but higher close to the equator. We have less SO4 in the free troposphere, but in SE Asia and the eastern equatorial Pacific Ocean we have considerably larger column burdens dominated by contributions in the lower troposphere. This is probably caused by our neglecting vertical exchange in cumulus clouds. Unfortunately BC columns are not available in the literature. The maxima in Europe and East Asia are dominated by fossil fuel combustion sources. The main contrast to SO4 is the low-latitude burdens to which biomass burning is the major contributor. Biomass burning would likely contribute more if emissions were injected at higher levels [Liousse et al., 1996].

Figure 3.

Integrated column burdens (μmol m−2) of annually averaged DMS, SO2, sulfate, and BC.

3.1.2. Seasonal Variability

[49] The seasonality in column burdens measured by a seasonality index, fs, defined and presented for a selection of areas in Table 2. The index varies between −2 for zero June–July–August values and 2 for zero December–January–February values, and fs = 2/3 (−2/3) when JJA (DJF) values are twice those in DJF (JJA). DMS has a winter maximum in the Northern Hemisphere, which is opposite to the results of Seland and Iversen [1999] where the DMS emission almost vanished over large sea areas in winter [Tarrason et al., 1995]. In the southern midlatitudes the DMS column is slightly larger in summer, but over Antarctica maximum occur in winter. The annual cycle is a combined result of biological production with a peak in spring, and the strength of boundary layer turbulence and sea surface waves, which in the extratropics peaks in winter.

Table 2. Seasonality Index fs = 2[CJJA − CDJF]/[CJJA + CDJF] for Column Burdens
 DMSSO2SO4BC
  1. a

    Range of values in selected regions.

NE America (80°–90°W, 30°–40°N)−1.5 to −10 to +1+1 to +1.5+0.2 to +0.7
Central Europe (5°–20°E, 45°–55°N)−1.5 to −1−0.5 to +0.5+0.5 to +0.70 to +0.5
SE Asia (100°–120°E, 15°–35°N)−1 to +1.5−1 to +0.2−0.5 to +1−0.9 to −0.5
Arctic−0.8 to −1.2−1 to −1.50 to +0.50 to +0.3
North Atlantic midlatitude (40°W, 40°N)−10 to +0.2+1+0.3 to +0.5
Tropical Atlantic−0.5 to +1−1 to +0.5−0.2 to +1.2−1.2 to −0.5
South Atlantic midlatitude (20°W, 40°S)−0.3 to −0.5+0.5 to +10+1.5
Antarctic+0.5 to +0.7−1 to −1.5−1.5+0.3 to +0.5

[50] For anthropogenic SO2 and SO4 the seasonal cycle is in the model only a consequence of chemistry and deposition. In reality also SOx emissions have a winter maximum in Europe, which would contribute to fs in a negative direction. The oxidant levels are considerably higher in summer than winter, but the efficiency of oxidation depends on availability of clouds. From the cloudiness numbers presented in Table 3, we conclude that the efficiency of wet-phase oxidation contributes significantly to SO2 variations as follows: (1) a summer column maximum in NE America, (2) no clear seasonality in Europe, and (3) only a slight winter maximum in SE Asia. In the Arctic winter, SO2 burdens are larger than those in summer, in accordance with seasonality in oxidation, dry deposition efficiency and meridional transport [Iversen, 1989]. However, the model predicts an Arctic-wide winter maximum of low-level cloudiness in contradiction to most observations in the region [e.g., Barrie, 1986; Rasch and Kristjánsson, 1998]. The SO2 levels in the Southern Hemisphere, and in the Antarctic in particular, are much lower than in the northern, and there are winter (JJA) SO2 maxima in areas influenced by anthropogenic emissions at the continents, and summer maxima in areas influenced by DMS, which is oxidized to SO2 and SO4 in the Antarctic summer.

Table 3. Modeled Annual Fractional Cover of Low Clouds (Clow) and Precipitation P (mm/d)
 Clow(JJA)Clow(DJF)P(JJA)P(DJF)
NE America (80°–90°W, 30°–40°N)0–10%20–50%1–31–2
Central Europe (5°–20°E, 45°–55°N)10–40%30–70%1–21–3
SE Asia (100°–120°E, 15°–35°N)20–50%20–50%0.5–50.5–1
Arctic80–92%90–97%0.7–1.20.4–1

[51] SO4 generally has summer maxima. In SE Asia the situation is strongly modulated by the phase and strength of the monsoon rain. Wet deposition in summer is efficient when SO4 production also is at its maximum. In the Arctic, a winter–spring maximum is expected [Iversen, 1989], but is not calculated. This model error is possibly connected with the wrong seasonal cycle of low-level Arctic cloudiness leading to exaggerated wet deposition. More studies are needed to understand this. In accordance with the work of Kasibhatla et al. [1997] there is a much larger seasonal SO4 burden cycle in North America than in Europe. This difference is potentially even larger due to the neglected European seasonal emission cycle.

[52] For BC the seasonality is pronounced in the Atlantic midlatitudes with maximum in June–August in both hemispheres. In the Southern Hemisphere the winter maximum is due to emissions from biomass burning. In central Africa, these emissions peak in January [Cooke et al., 1999]; hence negative fs values are seen in the tropical Atlantic Ocean. In all other areas seasonality is predominantly caused by variations in precipitation. This is particularly evident in SE Asia with the monsoon precipitation in summer (JJA) causing enhanced scavenging (see Table 3 for precipitation).

3.1.3. Production-Tagged Concentrations

[53] Figures 4 and 5 show the percentage of SO4 and BC column mass burdens that are tagged according to production mechanism. Figure 4 for SO4 can be compared with the work of Barth et al. [2000, Plate 3]. The a3 mode dominates except for over dry areas at low latitudes where the a1 mode is slightly more abundant. In the work of Barth et al. [2000] the aqueous phase was less dominating, in particular in remote areas where upper tropospheric contributions from gas-phase oxidation is important. Now, our a3 mode also contains contributions by coagulation, but the main reason for the difference is once again the neglect of transport in moist convective clouds. The maximum n-mode of ca. 5% is calculated in dry, subtropical regions. There is clearly no need for a separate budget for the a2 mode, which is produced by coagulation of the n-mode on accumulation mode.

Figure 4.

Relative contribution to annual column burdens of sulfate (%). SO4(n) = nucleation mode; SO4(a1) = accumulation mode by gas phase and condensation; SO4(a2) = accumulation mode by coagulation of SO4(n); SO4(a3) = accumulation mode by aqueous-phase oxidation and cloud-droplet coagulation.

Figure 5.

Relative contribution to annual column burdens of BC (%). C(n) = hydrophobic, externally mixed nucleation mode; Cx(a) = hydrophobic, externally mixed accumulation mode; C(a) = hydrophilic, internally mixed accumulation mode.

[54] For BC the hydrophilic accumulation mode dominates together with the hydrophobic nucleation mode that contributes mainly at low latitudes. The externally mixed hydrophobic Cx(a) mode has a percentage of more than 50 in the Arctic, even though only 10% of the fossil fuel BC originates in this mode. The coagulation of these particles with hydrophilic matter is much slower than for the much smaller n-mode particles. Agglomerate BC particles are abundant in Arctic haze [Rosen et al., 1981; Sheridan, 1989].

[55] Figure 6 shows that a major portion of the SO4 throughout the model atmosphere has been processed through the aqueous phase (a3 mode). SO4 produced in gas phase is mainly important in the lower troposphere at low latitudes. As noted above, this is likely caused by neglected transport in convective clouds, since we have much less gas-phase SO4 than Barth et al. [2000, Figure 15] had. SO2 that survives the convective scavenging will be injected into the upper troposphere where gas-phase oxidation will dominate due to low availability of liquid droplets. For the aqueous phase we have a factor 2–3 smaller absolute amounts in the upper troposphere (of the same reason), but elsewhere the patterns are similar. Figure 6 also shows that the model predicts considerable amounts of both hydrophilic and hydrophobic BC throughout the troposphere. Except in the tropics and the upper troposphere, hydrophobic BC is generally more abundant.

Figure 6.

Zonal averages of volume mixing ratios of the annually averaged sulfate (pmol (mol air)−1) produced in gas phase and aqueous phase and concentrations of hydrophobic and hydrophilic BC (ng(C) m−3).

3.2. Comparison With Measurements

[56] For models whose meteorology is not controlled by data assimilation, long series of are required to obtain statistical significance, and we only compare monthly, seasonal and longer averages preferably obtained from data over many years. For aircraft data we have to relax this principle if any comparison could be made. The observations in Europe cover periods of vastly changing emissions making multiyear averages difficult to interpret. Furthermore spatial representativity problems occur when point measurements are compared with model grid volumes. Finally, our calculated SO2 is not reduced from its value at the lowest model level to the level where measurements are taken (1 m). If c and c1 are concentrations at the lowermost model level and at 1m respectively, c1/c can be small (∼0.1) when turbulence and wind are weak. Averaged over a wide range of conditions, the calculated grid area SO2 should be larger than observed by a factor 1.2–2, provided the emission data are valid for the measurement period. The factor is particularly large at sites in major source regions where sites are biased toward “background” air to avoid local contamination. Despite these problems, the model should be able to well within one order of magnitude replicate differences between source regions and remote areas, and differences between the lower and upper troposphere.

3.2.1. Sulfur

[57] Figure 7 shows calculated versus observed ground-level SO2 and SO4 concentrations and sulfur wet deposition for a selection of sites in North America and Europe. The measurements for North America are averages for July 1988 to May 1990 and 1983–1992 for Europe. These data are those used by Kasibhatla et al. [1997, Figures 3–6] and Barth et al. [2000, Figures 3 and 4]. The IPCC [2001] emissions for 2000 for anthropogenic SOx are comparable to official data from European countries around 1990s [Vestreng and Støren, 2000]. In Europe emission reductions up to a factor 5 have occurred in some countries in the 1990s. The emissions for 2000 are also comparable to those used by Kasibhatla et al. [1997] and Barth et al. [2000].

Figure 7.

(a) Observed versus calculated yearly averaged SO2 and sulfate ground-level mixing ratios (ppb(v)) and sulfur wet deposition fluxes (mmol m−2) for winter (DJF) and summer (JJA) in eastern North America (EMEFS) [McNaughton and Vet, 1996]. Abscissa: observed; ordinate: calculated. Lines for perfect agreement, factor 2 and factor 10 errors are shown. (b) Same as Figure 7a, except for Europe (EMEP) [Schaug et al., 1987].

Figure 7.

(continued)

[58] Figure 7 shows that both in Europe and North America a major portion of the calculations are within a factor of two of observations for all components, and there is no general trend in error between low and high concentration regimes. In Europe there is a tendency to underestimate SO4 in winter and overestimate in summer (a factor 1.5 in either direction) partly because of missing seasonal variation in emissions. The larger overestimate for SO2 in summer in Europe (factor 3–4) are probably caused by the neglected vertical convective transport and inefficient scavenging in convective clouds. In North America summertime SO2 is a factor 2 larger than observed at the most polluted sites, while SO4 is underestimated around a factor 1.5. In addition to the lacking convective transport, oxidation may therefore still be too slow. Our SO4 calculations are closer to observations than the calculations of both Kasibhatla et al. [1997] and Barth et al. [2000]. In particular we have considerably more SO4 in winter, as a consequence of the assumed minimum in-cloud oxidation rate. Kasibhatla et al. [1997] showed that such oxidation favors increased SO4 levels in winter. The neglected vertical transport in cumulus clouds may contribute in the same direction but less efficiently than in summer.

[59] In Figure 8 we have compared our calculations of ground-level SO4 with annually averaged observations at remote oceanic as a function of latitude (data provided by J. Prospero and D. Savoie). Agreements are well within a factor 2 good at northern high and midlatitudes, and around a factor 2 with some scatter at southern high and midlatitudes. In the tropics, however, the model overestimates surface concentrations with an order of magnitude. We take this as a signature of the combined neglected vertical convective transport and inefficient wet deposition in cumulus clouds.

Figure 8.

Observed and calculated annual SO4 in remote oceanic areas (μg m−3). Observations obtained from Prospero and Savoie.

[60] Figure 9a shows that the model overestimates monthly DMS at Amsterdam Island and Cape Grim with a factor 1.5–3 with largest bias in late summer (February–March), yet the seasonal trend is reproduced. In contrast, Barth et al. [2000] overestimated DMS with a factor smaller than 2 and most in winter. Our larger DMS emissions are probably not sufficient to explain this difference. Our neglecting vertical transport in cumulus clouds in summer can be important as well, however, our nighttime oxidation to SO2 is also practically zero in the Southern Hemisphere. Chin et al. [1996] obtained close agreement with these observations even with higher effective emissions than ours, because they applied a hypothesized oxidation mechanism that produced SO2 more efficiently than their OH and NO3 oxidation.

Figure 9.

(a) Observed and calculated monthly DMS concentrations at Amsterdam Island and Cape Grim (ppt(v)). Observations taken from the work of Barth et al. [2000]. (b) Observed and calculated monthly SO2 concentrations at selected sites (ppt(v)). Observations taken from the work of Barth et al. [2000]. (c) Observed and calculated monthly concentrations of particulate sulfate in air (nmol (mol air)−1). Observations taken from the work of Barth et al. [2000]. The diagram for Mauna Loa are representative for the lower free troposphere. Two alternatives represent calculations, one below (695) and one above (598) the observation level.

Figure 9.

(continued)

Figure 9.

(continued)

[61] Figure 9b reiterates the conclusions from Figure 7 that SO2 is overestimated in summer in central Europe and eastern North America, and underestimated in winter in central Europe, and the explanations for this. In the Norwegian Arctic (Bear Island and Ny Ålesund) we obtain winter concentrations of SO2 well within a factor of 2, while the very small summer concentrations are underestimated. Also at Jergul, which is influenced by some very large point sources 150–200 km to the east, we get agreement for SO2 well within a factor 2 throughout the year.

[62] For SO4 (Figure 9c) we get a factor 3–10 more SO4 at the European sites compared to those obtained by Barth et al. [2000], while at the North American sites the factor is ca. 1.5. This brings our calculations in closer agreement with observations, but with some seasonal biases that was commented under Figure 7. Monthly SO4 levels in Iceland are well within a factor 1.5–2, as they also are at Jergul, but at Bear Island SO4 underestimates are more systematic, and between 1.5 and 2. At Ny Ålesund we underestimate winter concentrations about a factor 10, and at the two Canadian sites in the Arctic, SO4 is more than a factor 10 lower except in summer and autumn. Inefficient meridional transport may be part of the reason for this, but given that SO2 levels are well reproduced at Ny Ålesund, spurious wet scavenging of SO4 by low intensity but ubiquitous precipitation in low clouds spuriously predicted by the model during winter, could at least influence SO4 at Ny Ålesund.

[63] Free tropospheric observations from aircrafts are typically only available from short-lived campaigns. However, at the peak of Mauna Loa, 3400m above sea level, a long time series of SO4 representative for the lower free troposphere, is available and a model comparison is included in Figure 9c. We have taken out data from the nearest model levels above and below the observation height, and while data from the lowest level overestimates SO4 with a factor 3–4 in June, July, and August, data from the level above agree almost perfectly with the measurements. This illustrates the representativity problems when comparing point data with grid volumes.

[64] In Figure 10 we present concentrations of SO2 and SO4 for the periods and sites of the Pacific Exploratory Mission (PEM). The data were taken from the work of Barth et al. [2000, Figures 12 and 13]. Our free tropospheric concentrations of DMS are practically zero above η = 700, and we do not include the figure. The thick continuous lines in Figure 10 are model data from our basic run, other lines are for sensitivity tests described and commented in section 4.1. The effect of neglecting convective transport is most clearly seen for SO2 above η = 400 where we hardly calculate any concentrations, except in Japan (March) where calculations match within a factor 2 and at Easter Island (September) where we underestimate a factor 10. At low levels, we underestimate SO2 a factor 2 or more in Japan (March), there is a match within a factor 2 at Easter Island (September), Guayaquill (September), where Barth et al. [2000] overestimated a factor 2, and Hong Kong (February). We overestimate SO2 with a factor 2–5 at the remaining sites. Our largest low-level overestimate of SO2 is seen for Hong Kong (October), where also Barth et al. [2000] had a factor 2 overestimate. Between η = 800 and 500 we generally match observations within a factor 2.

Figure 10.

(a) Vertical profiles of SO2 compared to measurements (crosses) taken from the work of Barth et al. [2000]. Thick solid line: base run over 3 years; thin solid line: a repetition of base run for 1 year; dashed line: sensitivity test 4 with full convective transport; dash-dotted line: sensitivity test 5 with 10% convective transport. (b) Same as (a), but for vertical profiles of SO4.

Figure 10.

(continued)

[65] For SO4 the upper-level underestimates are not as evident as for SO2, rather we are well within a factor 2 above η = 600 in all cases, in contradiction to those obtained by Barth et al. [2000] who systematically overestimated SO4 a factor 2 and more at those levels. At low levels we overestimate a factor 2–5 at 4 sites, while there is little bias for other sites. We would advocate that our agreement with SO4 for these vertical profiles are better than in the work of Barth et al. [2000]. However, given our biases for SO2, we realize that this agreement probably is coincidental. Injecting volcanic SO2 at upper levels would help in some cases, but probably insufficiently. More research is needed to parameterize the transport and scavenging in convective clouds better.

3.2.2. Black Carbon

[66] Unfortunately, measurements of BC are even more sparse and sporadic than for sulfur. We have chosen a set of annually averaged measurements published by Cooke and Wilson [1996], Liousse et al. [1996], and Cooke et al. [1999]. Results are shown in Figure 11. There is a spread up of about a factor 10 between the sites in the agreement between observations and calculations, and on the average the model overestimates BC with a factor 1.6. Both this factor and the spread can be significantly influenced by low quality and misrepresentation of the measurement data. The data include continental and polluted rural sites, as well as remote oceanic and polar sites. They are sampled over very different periods, and their climatological representativity can be questioned. Nevertheless the calculations reproduce the measurements well within an order of magnitude for all concentration regimes, which was also found by Koch [2001]. Overestimated BC is found in tropical and subtropical oceanic areas, which we believe is linked to the neglected vertical transport in cumulus clouds and the low efficiency of convective wet scavenging.

Figure 11.

Observed and calculated air concentrations of BC (ng(C) m−3). (Top) scatterplot of annual averages; (bottom) annual averages in continental and oceanic areas as a function of latitude. Observations from the works of Cooke and Wilson [1996] and Liousse et al. [1996].

[67] Figure 12 shows a selection of sites with monthly data. BC at Amsterdam Island is mainly transported from southern Africa. Biomass burning peaks in winter/spring (Southern Hemisphere), which probably explains the modeled maximum. Measurements show typically a factor 4 smaller maxima. Cooke and Wilson [1996] overestimated BC at Amsterdam Island with a factor 2–3, while Liousse et al. [1996] underestimated slightly. Assuming that emission data are of good quality, the reason for the overestimate could be a too small hydrophilic fraction, too little precipitation, or too slow vertical transport. Sensitivity tests by Cooke and Wilson [1996] indicate that increasing the hydrophilic fraction is not sufficient to reduce our overestimates. From the hydrological CCM3 statistics of Hack et al. [1998], the precipitation in the Indian Ocean has no clear bias. Hence our assumption of ground-level biomass emissions probably is to blame. In comparison, Liousse et al. [1996] released biomass burning BC at 2 km height. At Mauna Loa we overestimate BC a factor 10 in December and January and in other months around a factor 2. We have an almost correct spring maximum in BC. The agreement is slightly better than in the work of Cooke et al. [1999]. At the South Pole underestimated BC of a factor 5–10 is seen in summer, in other parts of the year there is a close match. In the Arctic calculated BC is within a factor 1.5 in the summer–autumn, but more than a factor 10 underestimated in winter and spring, which is much the same results as calculated by Liousse et al. [1996] and Cooke et al. [1999]. The same behavior occurred for SO4, and we believe that the bias is linked to CCM3's overestimated low-level cloudiness in Arctic winter. At Mace Head we overestimate monthly BC throughout the year with a factor 3–10. Cooke et al. [1997] showed up to an order of magnitude variation in the BC levels depending on the origin of the air masses. We have crudely simulated this situation by comparing BC under influence of continental air with calculations in a land grid point, while measurements under marine influence are compared with a nearby marine point. The calculations nevertheless overestimate measured BC around a factor 5 in both cases. Except in winter, Koch [2001] did not calculate a similar BC bias for Mace Head.

Figure 12.

Observed and calculated monthly concentrations of BC in air (ng(C) m−3). Observations taken from the works of Cooke and Wilson [1996], Liousse et al. [1996], and Cooke et al. [1999].

3.3. Scenario for Year 2100 (IPCC SRES A2)

[68] We have made a 5-year scenario run for 2100 for which the emissions of BC and anthropogenic SOx constitute the only difference from the run for 2000. The greenhouse forcing, sea surface temperature, and the oxidant levels are unchanged, hence this experiment is more like a sensitivity test for different emissions rather than a true climate scenario. Again we only use output for the last 3 of the 5 years. Figure 13 shows column burdens for SO4 and BC for emission scenarios 2000 and 2100. For SO4 the column burden is projected to decrease over major parts of the Northern Hemisphere, in particular in the Arctic and in Europe. In the Southern Hemisphere the changes are small, except for in Africa south of the equator and downwind in the Atlantic Ocean where column burdens are doubled. For BC the projected changes are considerably larger, with column burdens increased with a factor 2–2.5. The only exception is over areas south of 50°S where there are negligible changes. If such a situation will hold true, the direct global radiative forcing of anthropogenic aerosols may turn from negative to positive due to the absorption caused by BC.

Figure 13.

Calculated column burdens of sulfate (mg(SO4) m−2) and BC (mg(C) m−2) in 2000 (upper) and 2100 (lower).

3.4. Budgets

[69] In our model 34% of the emitted DMS removed and does not contribute to any DMS in the model atmosphere. The emission and lifetime for DMS available for conversion to SOx are therefore those given in parentheses in Table 4. These are the same to those of Feichter et al. [1996] who used a MSA yield of zero. Our total lifetime for DMS is the same as those of Rasch et al. [2000a], but this is influenced by the immediate removal of DMS as MSA in our model. As stated by Rasch et al. [2000a] MSA is a product of OH oxidation, hence we have slower resolved oxidation of DMS than Rasch et al. [2000a]. We only calculate the nighttime reaction over Northern Hemispheric continents close to the ground. Given our higher DMS emissions in the North Atlantic Ocean, one reason for our slower rate could be missing turnover by the nitrate radical in the Arctic winter. The spread among several of the models is probably caused by different oxidation rates. Note that Restad et al. [1998] use the same OH fields as we do, but they neglect the nighttime reaction with NO3. They obtain more than twice our lifetime, realizing that this most probably is an overestimate.

Table 4. Emissions, Burdens, Turnover Times, and SO2 Yield for DMS
 This workRasch et al. [2000a]Langner and Rodhe [1991]Pham et al. [1995]Feichter et al. [1996]Chin et al. [1996]Lelieveld et al. [1997]Restad et al. [1998]Koch et al. [1999]Chin et al. [2000]
  • a

    Numbers obtained when the fraction (34%) assumed to produce MSA is excluded from the DMS emissions.

Source (Tg(S)/a)26/17a161620172216161113
Burden (Tg(S))0.100.060.130.050.100.060.020.220.060.07
Lifetime (days)1.4/2.2a1.430.92.21.00.55.01.92.0
SO2 Yield (Tg (S)/a)17141618172216161012

[70] Table 5 shows that the lifetime of SO2 is practically unchanged from 2000 to 2100. For SO4 it is 15% longer in 2100, and the model's production efficiency for global SO4 (P in Table 5) is higher. The anthropogenic emissions in 2100 are less exposed to clouds, and oxidation of SO2 is slower and wet scavenging of SO4 less efficient. The importance of both the height and geographical region of emission sources is seen by comparing the total budget numbers with those for natural sulfur. Natural SO2 is predominantly produced from ground-level DMS over oceans, and is exposed to efficient dry and wet deposition, as well as in-cloud oxidation to SO4, and the lifetime is only slightly more than 1 day. Also the SO4 production efficiency is smaller. Our turnover time for SO2 is in the mainstream among the models, but the SO4 production efficiency is lower than in the model of Rasch et al. [2000a] and compared to the average. The SO2 turnover time is exactly the same as obtained by Feichter et al. [1996] but smaller than obtained by Rasch et al. [2000a]. These properties are probably caused by our neglected vertical transport and inefficient wet deposition in cumulus clouds, since we have more SO2 and SO4 in the lower levels where deposition is fast. The very short lifetime estimated by Pham et al. [1995] is due to no oxidant limitation. Rasch et al. [2000a] and Koch et al. [1999] counted SO2 oxidized to SO4 in precipitation and subsequently deposited, as a SO4 source. In traditional terms their SO4 lifetime is therefore slightly longer than given in the table, and our lifetime is shorter than theirs, as well as most of the other models.

Table 5. Budget Parameters for the Production of SO4 Particles in the Atmosphere
 Sulfur emission (Tg(S)/a)SOx source (Tg(S)/a)SO2 deposition (%)Aqueous SO4 production (%)Gaseous SO4 production (%)SO2 burden (Tg(S))T(SO2) (days)SO4 source (Tg(S)/a)SO4 wet deposition (%)SO4 burden (Tg(S))T(SO4) (days)P (days)
  • a

    T are turnover times, P = sulfate burden per daily SOx (SO2 + SO4) emission = SO4 production efficiency.

  • a

    Rasch et al. [2000a] and Koch et al. [1999] assign SO2 oxidized in precipitation as produced and deposited sulfate.

This work
  2000 total99.990.440.844.413.30.371.553.8830.523.52.1
  2100 total91.282.340.841.816.00.361.648.7810.544.02.4
Natural30.922.046.445.18.50.071.112.1800.113.31.8
Other works
Langner and Rodhe [1991]98.098.045.642.880.301.253.5840.774.62.9
Pham et al. [1995]124.6122.849.545.25.30.20.662.0730.84.72.3
Feichter et al. [1996]100.7100.749.133.717.20.431.551.3870.634.52.3
Chin et al. [1996]96.795.648.643.5 7.80.341.349.1890.533.92.0
Chuang et al. [1997]106.0106.054.339.95.80.361.248.4890.554.11.9
Lelieveld et al. [1997]94942356140.62.372751.15.34.3
Restad et al. [1998]91.991.94543120.422.050.7830.624.52.5
Koch et al. [1999]83.082.343.4a38.4a15.90.562.646.6a80.3a0.735.7a3.2
Chin et al. [2000]93.992.556.626.515.10.431.840.785.30.635.82.5
Rasch et al. [2000a]838132a55a120.41.955a93a0.604.0a2.7

[71] The turnover time for BC (Table 6) is closer to those of Liousse et al. [1996] and Koch [2001] (“case S”) than to Cooke and Wilson [1996]. The number in the work of Cooke et al. [1999] is only fossil fuel BC, and our corresponding number for 2000 is 3.8 days. This is due to a swift transfer from hydrophobic to hydrophilic particles brought about by coagulation. According to Koch [2001] this transfer is even faster even though only exposure to SO4 produced in gas phase turns BC hydrophilic. Again this is a consequence of a quick transport to the upper tropospheric levels where gas-phase SO4 is abundant. Table 6 also reveals a different behavior for BC from biomass burning than from fossil fuel. The efficient production of hydrophilic BC (the large P-value) is caused by the fact that 50% of the biomass burning BC is emitted as hydrophilic. Since this BC is mainly emitted in the tropics, however, it and can be swiftly dispersed to subtropical areas where clouds are rare and removal slow.

Table 6. Budget Parameters for the Production of BC Particles in the Atmosphere
 Hydrophobic BCTotal BCPphilic (days)
Source (Tg(C)/a)Burden (Tg(C))T (days)Source (Tg(C)/a)Burden (Tg(C))T (days)
  1. a

    Pphilic = burden of hydrophilic BC per total daily BC emission = hydrophilic BC production efficiency.

This work
  2000 total10.50.072.412.40.164.72.7
  2000 biomass burning1.90.023.43.80.076.44.7
  2100 total24.50.162.428.80.374.72.7
Cooke and Wilson [1996] (1)14.014.00.379.8
Cooke and Wilson [1996] (2)14.014.00.307.9
Cooke and Wilson [1996] (3)14.014.00.215.6
Liousse et al. [1996]00indefinite12.30.133.93.9
Cooke et al. [1999] (fossil fuel)5.10.075.3
Koch [2001] (case S)12.40.041.0912.40.154.43.2

4. Discussion

4.1. Sensitivity Tests

[72] The properties of the model rely with the actual choice of parameters for physical parameters. As explained in the introduction, climate models cannot presently afford to resolve aerosol and cloud processes from first principles. Parameterizations involving poorly determined parameter values are unfortunately necessary. Here we have conducted 6 sensitivity experiments to investigate the impact of some of the uncertain model formulations, partly inspired by similar tests by Seland and Iversen [1999].

[73] The tests all start from the same global distributions of sulfur and BC components and last 13 months. Statistics are calculated for the last 12 months. The basic run is repeated under the same conditions to facilitate comparison with the test runs. Table 7 defines the 6 tests, and results are seen in Table 8.

Table 7. 1-Year Sensitivity Experiments, All Starting From the Same Initial Conditions
BasicControl experiment, IPCC [2001] emissions
Test 1Temperature limit for pure ice clouds changed from −25°C to −40°C
Test 2Below-cloud scavenging rate for particles divided by 10
Test 3Minimum wet-phase oxidation rate divided by 10
Test 4Full vertical mass flux transport in cumulus clouds
Test 510% vertical mass flux transport in cumulus clouds
Test 6In-cloud scavenging of particles by convective precipitation in entire grid volumes
Table 8. Burdens and Turnover Times for SO2, Sulfate, and BC for the Sensitivity Experiments
 SO2 burden (Tg(S))T(SO2) (days)SO4 productionSO4 source (Tg(S)/a)SO4 wet deposition (%)SO4 burden (Tg(S))T(SO4) (days)P (days)BC burden (Tg(C))T(BC) (days)Pphilic (days)
Effective rate (10−6 s−1)In-cloud (%)
Basic0.371.54.47654830.523.52.10.164.72.7
Test 10.351.44.77755830.483.21.90.154.52.4
Test 20.371.54.47654790.604.02.40.206.03.3
Test 30.381.64.27654830.523.52.10.164.72.7
Test 40.482.03.97461871.559.36.30.319.26.3
Test 50.391.64.47656840.704.62.80.185.53.3
Test 60.371.54.47656900.261.71.00.102.81.4

[74] Test 1 investigates the effect of allowing mixed-phase clouds down to −40°C instead of only −25°C. Chemistry and scavenging is reduced linearly with temperature from 0°C to the temperature of pure ice clouds, and the test allows more liquid water. Also the solubility of H2O2 increases with lower temperature causing higher in-cloud oxidation rates for SO2. The result is more SO2 oxidation and more scavenging of SO2, SO4, and hydrophilic BC. Column burdens are reduced with up to a factor 2.5 for SO2 and a factor 2 for SO4 and BC in the Arctic and Antarctic. Over midlatitude source regions reductions are typically from 5% to 15%, while at low latitudes the reduction is small. The relative reductions in SO4 and BC are typically twice as large in the free troposphere as in the boundary layer. The overall turnover times for all three components are reduced with less than 10%. All in all, the increased availability of liquid water increases the wet scavenging. The increased in-cloud SO2 oxidation produced SO4 more susceptible to scavenging.

[75] Test 2 addresses the considerable uncertainty in the quantification of below-cloud scavenging of particles. In our basic run we have chosen to use a quite efficient rate based on the work of Hobbs [1993]. If instead data presented by Seinfeld and Pandis [1998, pp. 1020–1026] were used, the rate would be much smaller. With 10 times smaller below-cloud scavenging rates for all particles we obtain 40% longer turnover times for BC and a 15% increase for SO4. Since BC consists of primary particles while SO4 particles are secondary, larger amounts of BC than SO4 reside below clouds in emission areas. In emission areas the increase of BC column burdens are 10–25%, while for SO4 it is 3–5%. Hence relatively more BC than SO4 becomes available for transport to remote areas. Since BC also has a hydrophobic component, the increase in its burden relative to SO4 in remote areas is further enhanced. Both BC and SO4 increase their burdens with up to 50% in the tropics, while in remote high latitudes BC increases up to 150% and SO4 only a little more than 5%. The production efficiency of hydrophilic BC is increased by 20% and of SO4 by 15%.

[76] Test 3 investigates the sensitivity to the quite arbitrarily chosen minimum in-cloud nonphotochemical oxidation rate for SO2. Kasibhatla et al. [1997] used a rate for an unidentified nonphotochemical oxidation pathway of 10−6 s−1 in winter and twice as high in summer. When clouds are present our value of 5.6 × 10−6 s−1 is considerably larger, hence the test uses a 10 times smaller value. BC is not influenced, and the annual SO4 column is only slightly changed with a decrease up to 5% near the largest European emissions. The increased SO2 is transported away from the sources, and the largest relative increase of up to 25% is obtained in the Arctic. This increase in SO2 levels also produces a slight increase in remote SO4 columns of 2–4%, but the SO4 production efficiency (P) is unaltered. Hence, our budgets do not depend vitally on this oxidation pathway.

[77] In test 4 we use full vertical fluxes in convective clouds for transport of all airborne contaminants. Table 8 reveals very large effects. The turnover time for SO2 is increased from 1.5 to 2 days, for BC it is almost doubled, and for SO4 it is increased by a factor 2.7. Hydrophilic BC production becomes 2.5 times as efficient and the production efficiency for SO4 is tripled. The SO2 column is decreased by up to 10% in source areas, but otherwise it is typically increased 20–50% over continents and up to 200–400% over oceans. Largest relative increase is seen over subtropical ocean areas downwind (west) of continents, as a consequence of deep tropical convection and the associated divergent flow in the upper troposphere and the easterly winds in the tropical troposphere. Both SO4 and BC increase their column burdens in all points. The relative increase in source areas is 20–100% for BC and 50–150% for SO4. In remote regions it is typically more than 200% with a maximum up to 700–900% in the southern subtropics.

[78] The dashed lines in Figure 10 show results from this test to be compared with PEM observations. Main changes for SO2 is seen as higher concentrations in the upper free troposphere, while SO4 has a close to uniform profile at most sites with an order of magnitude overestimate in the upper levels. Figure 14 shows zonal concentration averages. The figure should be compared with Figure 1 of Barth et al. [2000] and our own Figure 1. The burden for all components increases more than an order of magnitude over huge parts in the free troposphere, while surface concentrations only decrease slightly, and mainly at low latitudes. The pattern of our results for DMS, SO2, and SO4 is quite comparable to the results of Barth et al. [2000], but both SOx components have considerably higher burdens except in the lower Arctic troposphere. The emissions are different, but the main reason for deviations from the work of Barth et al. [2000] is our decreased wet scavenging rates for convective clouds. For BC the original results agreed well with the results of Liousse et al. [1996, Figure 13b], hence test 4 yields much higher BC concentrations, more reminiscent of those obtained by Koch [2001] in the test without convective scavenging.

Figure 14.

Zonal averages of volume mixing ratios of the annually averaged DMS, SO2, sulfate (pmol (mol air)−1), and BC (ng(C) m−3) in the case of full vertical transport in convective clouds (test 4).

[79] Test 5 investigates the effects of using only 10% of the convective mass fluxes for transport of sulfur and BC. In the real atmosphere there are ample possibilities for air in an updraft plume to be exchanged with ambient air by turbulence without changing its total mass. This test is a crude attempt at reducing the convective transport to a possibly more reasonable level than obtained when no exchange takes place below the plume top. The test yields a 30% increase in turnover times for SO4 15% for BC, a 5% for SO2. The production efficiency for SO4 is increased 30%, and for hydrophilic BC 20%. Increases in column burdens for all components compared to the base case are up to 100% in subtropical areas and typically 10–20% in source regions. Figure 15 shows the new zonal averaged concentrations. The upper tropospheric pattern for DMS in the tropics, indicates that 10% mass fluxes possibly is insufficient. In Figure 10 the dash-dotted lines are associated vertical profiles for comparison with PEM. Judging from the SO2 profiles, 10% mass fluxes are too small, but from the SO4 results it is rather the opposite, even though agreements with free tropospheric observations are mostly within a factor of 2. We believe the reason for this difference is inefficient convective cloud scavenging.

Figure 15.

Same as Figure 14, but in the case of 10% vertical transport in convective clouds (test 5).

[80] In test 6 we therefore have increased the effective convective scavenging rate. Here we assume that the entire grid volume is exposed to precipitation whenever there is a convective cloud present. The rationale is that the cloud-producing convective motion causes any air parcel in the column below the cloud top to spend a considerable time inside the cloud over a time step. This is probably too efficient for air well above the boundary layer, but since most of the contaminants exist in the lower portion of the troposphere the test is sensible. SO2 is not included in the test. SO4 and BC decrease in column burdens up to 80% in the tropics (100% meaning nothing is left). Elsewhere, the reduction is typically from 20% to 40%. The turnover times are reduced to half its original value for SO4 and by 40% for BC, and the production efficiency for SO4 and hydrophilic BC are halved. These drastic consequences are caused by the higher concentrations at low levels due to the neglected transport in convective clouds. Hence, these effects must be treated consistently.

4.2. Intercontinental Transport

[81] We have selected three major emission regions for allocation of contributions. The regions are North America, Europe and Asia, and are dominated by anthropogenic, fossil fuel emissions for SOx and BC. Separate 2-year runs have been made with anthropogenic emissions only for each source region. Table 9 shows the burdens and turnover times for the three emission regions based on results from year 2. Europe is the northernmost of the regions and SO2 emitted there is therefore less efficiently oxidized to SO4, while Asian SO2 is more swiftly oxidized than North American due to more clouds. Europe also stands out with the shortest particle turnover times and smallest SO4 and hydrophilic BC production efficiencies (P), due to frequent precipitation. Consequently, even though Europe emits 21% of the total sulfur, only 14% of the SO4 burden originates from Europe. For North America the numbers are 13% and 12% respectively, and for Asia 31% and 27%. Similar results were presented by Rasch et al. [2000a], except for Asian sulfur, which in parts is strongly influenced by the vertical transport in tropical cumulus clouds.

Table 9. Burdens and Turnover Times for SO2, SO4, and BC Allocated to Three Source Regions
 SO2 burden (Tg(S))T(SO2) (days)SO4 productionSO4 source (Tg(S)/a)SO4 wet deposition (%)SO4 burden (Tg(S))T(SO4) (days)P (days)BC burden (Tg(C))T(BC) (days)Pphilic (days)
Effective rate (10−6 s−1)In-cloud (%)
Europe0.0741.63.9859.8840.0752.81.50.0313.41.6
North America0.0471.64.4736.9850.0623.32.00.0054.12.1
Asia0.101.55.17917.0860.143.01.80.0404.01.8

[82] Even though the efficient deposition of European emissions reduces Europe's contributions to column burdens, its relative contribution to global acid deposition is not reduced. Figure 16 shows maps of sulfur deposition from North American, European and Asian sources in the Northern Hemisphere. The maps for North American and European depositions can be compared with the maps of Tarrason and Iversen [1998], since the emissions were almost the same. Except for the contribution in the tropical Pacific Ocean, the patterns are quite similar. Deposition of North American origins penetrates 500–1000 km further into Europe than predicted by Tarrason and Iversen [1998]. Considerable amounts of Asian sulfur may also contribute to acidity west of the Rocky Mountains, and possibly also on the east coast by transarctic transport.

Figure 16.

Annually accumulated deposition of sulfur from three source regions (mg(S) m−2) (isolines 5, 10, 25, 100, and 500).

5. Conclusions

[83] Experiments have been made with a semiparameterized scheme for sulfur and BC implemented in the NCAR CCM3. The scheme tags particle concentrations according to production, allowing a posteriori calculations of size distributions, optical properties, and properties relevant for cloud physics. By omitting vertical transport in cumulus clouds we obtain within a factor 2 agreement with ground-level observations for sulfur over wide areas, and for SO4 in parts of the free troposphere. For BC agreement is well within a factor 10 at ground level in many regions. However, SO4 and BC are overestimated by a factor 10 at ground level at low latitudes, and SO2 is largely missing at upper tropospheric levels. This is most certainly due to the neglected vertical transport in convective clouds. For both SO4 and BC we underestimate concentrations in the Arctic. The reason may be linked to exaggerated low-level winter cloudiness and precipitation modeled in the Arctic, but this remains uncertain. Turnover times for SO2 are similar to, and for SO4 slightly shorter than other published results. For BC turnover times are comparable to but slightly longer than obtained by Liousse et al. [1996] and Koch [2001].

[84] We have focused on cloud-related issues for aerosol generation, transport and deposition. Cloud schemes in general do not provide information on the fraction of time an air parcel in a grid box is cloudy or exposed to precipitation, and neither is the expected Lagrangian cloudy (or clear air) residence time for grid box air parcels parameterized. These are crucial parameters for estimating efficient oxidation and scavenging rates from grid volume averaged quantities.

[85] Sensitivity tests show that atmospheric burdens of SO4 and BC depend heavily on the parameterization of vertical transport and scavenging in deep convective clouds. These processes must be parameterized consistently, otherwise factor 2 errors in global burdens may arise, and vertical profiles will be wrong. Sensitivity of 10–20% for global burdens is estimated depending on how scavenging and chemistry in mixed-phase clouds are treated, and on chosen below-cloud scavenging parameters for particles. When detrainment of air in modeled convective updrafts entirely takes place at the plume tops, boundary layer air is efficiently transferred to the model's upper troposphere. Parameterizations should be reevaluated with a view to also model contaminant processes. Exchange of air between updraft and downdraft convective plumes and with ambient clear, should be estimated as an integrated part of cumulus parameterization schemes in climate models.

Acknowledgments

[86] This paper is part of the RegClim project financed by the Research Council of Norway. The work has also received support from the Research Council's Programme for Supercomputing through grants for computer time. We are grateful to P. J. Rasch for providing an extended version of the CCM3 code and for numerous discussions. Emission data for the IPCC scenarios were provided by J. Penner. Cooperation and discussions with J. E. Kristjansson and A. Kirkevåg are gratefully acknowledged.

Ancillary