We use observations of the mass-independent oxygen isotopic composition (Δ17O) of sulfate in the marine boundary layer (MBL) to quantify the sulfate source from aqueous SO2 (S(IV)) oxidation by O3 in alkaline sea-salt aerosols. Oxidation by O3 imparts a large Δ17O signature to the resulting sulfate (8.8‰) relative to oxidation by H2O2 (0.9‰) or by OH or O2 (0‰). Ship data from two Indian Ocean Experiment (INDOEX) cruises in the Indian Ocean indicate Δ17O values usually <1‰ in the submicron sulfate aerosol but considerable variability in the supermicron sulfate with frequent occurrences above 1‰ and up to 6.7‰. The large Δ17O values are associated with high concentrations of sea-salt aerosols, providing evidence for the S(IV) + O3 pathway. We use a global chemical transport model (GEOS-CHEM) to interpret quantitatively the INDOEX observations and to assess the global importance of sulfate production in sea-salt aerosols. The model accounts for titration of sea-salt alkalinity in the MBL by uptake of acid gases (SO2, H2SO4, and HNO3), shutting down the S(IV) + O3 pathway. We find that this titration occurs rapidly over much of the oceans except at high latitudes (strong sea-salt emission) and is due to both the S(IV) + O3 reaction and HNO3 (g) condensation; that is, sulfate formation in sea-salt aerosols is limited by the alkalinity flux from the ocean and by competition for this alkalinity supply from HNO3 (g). The model is consistent with the Δ17O magnitudes and patterns in the INDOEX data. Titration of alkalinity is critical for the success of the model simulation. Regeneration of sea-salt aerosol alkalinity by OH uptake is inconsistent with the Δ17O observations in INDOEX. Model results indicate that sulfate production in sea-salt aerosols decreases MBL SO2 concentrations and gas phase H2SO4 production rates by typically 10–30% (up to >70%) and increases MBL sulfate concentrations by typically >10% (up to 30%). Globally, this mechanism contributes 9% of atmospheric sulfate production and 1% of the sulfate burden. The impact on H2SO4 (g) formation and implications for the potential formation of new particles in the MBL warrants inclusion in models examining the radiative effects of sulfate aerosols.
 Sulfate is a major component of the atmospheric aerosol and drives the formation of new aerosol particles through nucleation. The implications for scattering of solar radiation and for cloud microphysics represent one of the largest uncertainties in current assessments of climate change [Intergovernmental Panel on Climate Change (IPCC), 2001]. Sulfate is mainly produced within the atmosphere by oxidation of SO2, which is itself directly emitted (fossil fuel combustion, industrial processes, and volcanoes) or produced within the atmosphere by oxidation of reduced sulfur species such as dimethyl sulfide (DMS) emitted by oceanic phytoplankton. The oxidation of SO2 can take place in the gas phase [Stockwell and Calvert, 1983] and in the aqueous phase [Schwartz, 1987]. The branching ratio between these two pathways is not well understood but plays a critical role in determining the nucleation potential and optical properties of the sulfate aerosol.
 Oxidation in the aqueous phase is thought to occur primarily in clouds because of their high liquid water content [Schwartz, 1987]. However, deliquescent sea-salt and dust aerosols have also been suggested as important sites for aqueous phase sulfate production [Chameides and Stelson, 1992; Sievering et al., 1991; Song and Carmichael, 2001] because of the rapid rate of SO2 oxidation by ozone in alkaline solutions. Production in sea-salt aerosols could greatly affect sulfate concentrations and size distributions in the marine boundary layer (MBL) [Song and Carmichael, 2001], with consequences for both direct radiative forcing [Shaw, 1983] and indirect radiative forcing via cloud formation [Charlson et al., 1987]. Here we exploit oxygen isotope measurements of sulfate in the MBL to quantify production in sea-salt aerosols, and we interpret the results in terms of sources and titration of MBL alkalinity.
 Gas phase oxidation of SO2,
produces H2SO4 (g) that can nucleate new particles under favorable conditions, increasing aerosol number density and the population of cloud condensation nuclei (CCN) [Andronache et al., 1997; Kumala et al., 2000; Weber et al., 2001]. Sulfate produced in the aqueous phase is present in larger particles and does not lead to nucleation of new particles. Aqueous phase sulfate formation involves dissolution of SO2 followed by acid-base dissociation of SO2 · H2O to HSO3− (pKa1 = 1.9) and SO32− (pKa2 = 7.2). We refer to total dissolved SO2 as S(IV) ≡ SO2 · H2O + HSO3− + SO32−. Oxidation of S(IV) takes place by dissolved H2O2,
 The alkalinity budget of the MBL is thus critical for quantifying the importance of S(IV) oxidation in sea-salt particles. Once the alkalinity is consumed, the pH of the aerosols drops rapidly, and (R3) effectively shuts off [Chameides and Stelson, 1992]. A schematic for the processes involved is shown in Figure 1. Alkalinity supplied by the emitted sea salt is consumed not only by (R3) but also by condensation of gas phase acids (H2SO4, HNO3, and RCOOH) [Davies and Cox, 1998; Ten Brink, 1998; Katoshevski et al., 1999; Song and Carmichael, 1999; Guimbaud et al., 2002]. Coagulation of acid sulfate particles would also consume sea-salt aerosol alkalinity, but it is negligibly slow relative to condensation of acid gases [Gong and Barrie, 2003]. Ammonia emitted from the ocean and subsiding from the free troposphere is an additional source of alkalinity to the MBL; however, it is not significantly taken up by droplets with pH > 6 [Jacob et al., 1986] and therefore cannot sustain sea-salt aerosol alkalinity to allow (R3) to proceed.
 Alkalinity sources in excess of bulk seawater alkalinity have been proposed to impact the sulfur cycle in the MBL. Sea-salt aerosol measurements of excess calcium are attributed to a biogenic source of alkalinity (CaCO3 produced by coccolithophore phytoplankton), which would increase the estimated alkalinity flux from the ocean surface, particularly in regions of high oceanic productivity [Sievering et al., 2004]. On the basis of laboratory experiments, Laskin et al.  suggested that the reaction of the hydroxyl radical (OH) with deliquesced NaCl particles could continuously supply alkalinity through OH− formation at a rate comparable to sulfate formation via (R3). However, the atmospheric implications of Laskin et al.'s results have been challenged because of their neglect of gas phase diffusion limitation for OH uptake [Sander et al., 2004] and acidification by acids other than SO2 [Keene and Pszenny, 2004]. As we will see, the observed oxygen isotopic composition of sulfate argues against a continuous supply of alkalinity from the NaCl + OH reaction.
 We use here ship measurements of the oxygen isotopic composition of sulfate over the Indian Ocean from the Indian Ocean Experiment (INDOEX) campaign [Ramanathan et al., 2001] to gain information on sulfate formation pathways in the MBL. The different oxidants (O3, H2O2, and OH) have distinct oxygen isotopic compositions that are passed to sulfate upon oxidation [Savarino et al., 2000]. Constraints from oxygen isotopes have been used previously to infer the formation pathways of sulfate in continental air [Lee et al., 2001; Lee and Thiemens, 2001] and in ice cores [Alexander et al., 2002, 2004]. However, quantitative interpretation is difficult on the basis of measurements alone. We show here that a global model simulation including oxygen isotopes can help for that purpose and can also allow for extrapolation to the global scale.
2. Isotopic Constraints on Sulfur Oxidation Pathways
 Kinetic and equilibrium processes such as isotopic exchange or diffusion fractionate oxygen isotopes according to the following relation [Matsuhisa et al., 1978]:
where Rx = [xO]/[16O] is the isotopic ratio (x = 17 or 18), measured relative to an international reference standard (R0x). The standard used for oxygen is standard mean ocean water. The relationship (1) is termed mass-dependent fractionation. Any process that does not follow (1) is termed mass-independent and is characterized by the Δ17O value where [Farquhar et al., 2000]
For mass-dependent fractionation, Δ17O = 0‰. Atmospheric H2O and O2 have Δ17O at or near 0‰ [Thiemens, 1999]. SO2 has Δ17O = 0‰ throughout the troposphere because of isotopic exchange with water vapor [Holt et al., 1981], which erases any source-derived isotopic signature.
 Laboratory experiments show that a positive Δ17O in sulfate originates from aqueous phase oxidation of SO2 by H2O2 and O3 through simple transfer of the isotopic anomaly from the oxidant to the sulfate product following (R2) [Savarino et al., 2000],
Other sulfate sources including gas phase oxidation by OH in the troposphere, metal-catalyzed oxidation by O2, and primary sulfate are mass-dependently fractionated (Δ17O = 0) [Savarino et al., 2000; Lee et al., 2002]. Since S(IV) oxidation by O3(R3) is the only mechanism producing sulfate Δ17O values >1‰, observations of atmospheric sulfate Δ17O values >1‰ are a robust indicator of the importance of O3 in sulfate formation. Sulfate in the atmosphere does not undergo further isotopic exchange.
 The Δ17O values of sulfate formed in fresh sea-salt aerosols should be determined by O3 oxidation and should have Δ17O = 8.8‰, while Δ17O values of sulfate formed in clouds should be <1‰ because of low cloud water pH. Rainwater pH during the two INDOEX cruises discussed in this work ranged between 3.7 and 5.6 [Kulshrestha et al., 1999], sufficiently low to suppress S(IV) oxidation by O3. Δ17O measurements for MBL sulfate thus provide a means to distinguish between S(IV) oxidation in sea-salt aerosols versus other processes.
3. Ship Cruise Data
 A high-volume air sampler with a four-stage cascade impactor was operated on board the scientific vessel OR/V Sagar Kanya during its 120th cruise (cruise 120) from 27 December 1996 to 31 January 1997 as part of the pre-INDOEX 1997 campaign and during cruise 133 from 18 February to 31 March 1998 as part of the INDOEX 1998 first field phase. The high-volume air sampler was installed in the front of the ship to minimize contamination from the ship exhaust. Cruise tracks are shown in Figure 2. They covered the Indian Ocean from about 15°N to 15°S latitude and from 60° to 79°E longitude. Glass filter samples were collected over 48-hour intervals. Immediately after collection, the filters were placed inside a sealed plastic bag and were refrigerated. A total of 14 collections were made during pre-INDOEX 1997, and 12 collections were made during INDOEX 1998, covering the course of each cruise. After each cruise the filters were sent to La Jolla, California, for isotopic analysis.
 The collected sulfate was divided into two groups according to particle size: fine (≤1 μm) and coarse (>1 μm). The filters were submerged in a beaker with 100 mL of Millipore water and were placed in an ultrasonic bath for 15 min to dissolve sulfate ions. Any insoluble matter was filtered from the solution with 0.22 μm pore size cellulose filters. Activated carbon was added to the solution and was filtered to eliminate dissolved organic species.
 For the pre-INDOEX 1997 samples, sulfate was extracted from solution through precipitation as barite (BaSO4) using BaCl2, followed by conversion of BaSO4 to CO2 and fluorination (BrF5) of CO2 to yield O2 suitable for oxygen isotopic analysis as described by Bhattacharya and Thiemens . The precision of these Δ17O measurements is ±0.1‰. For the INDOEX 1998 samples we used a more recently developed technique [Savarino et al., 2001; Alexander et al., 2002]. Briefly, the sulfate ions were separated by ion chromatography, were converted to Ag2SO4 using a cation exchange column, and were pyrolyzed at 1000°C in a continuous flow system to yield O2 suitable for oxygen isotopic analysis. The precision of the Δ17O measurements with this method is ±0.3‰.
 The isotopic compositions reported in this paper have been corrected for primary sea-salt sulfate (ssSO42−) using Na+ concentration measurements, according to the relation ([ssSO42−]/[Na+])sample = ([SO42−]/[Na+])seawater = 0.252 mol mol−1 [Calhoun et al., 1991]. Subtraction of ssSO42− from the sulfate measurements yields non-sea-salt sulfate (nssSO42−). For the INDOEX 1998 cruise, [Na+] was measured directly in our samples by ion chromatography (Dionex 4500). For the pre-INDOEX 1997 cruise we used [Na+] data from a bulk aerosol sampler (D. L. Savoie, personal communication, 2004) running continuously next to our sampler, and we assumed a 10/90 fine/coarse mass fraction for sea-salt aerosols [Gong et al., 2002].
Figure 3 shows the nssSO42− Δ17O (‰) and Na+ concentration (μg m−3) measurements plotted against latitude. The pre-INDOEX 1997 Δ17O values are <1‰ except for the coarse nssSO42− south of the equator which displays large values, up to 6.7‰. A similar latitudinal gradient is observed for [Na+]. During the INDOEX 1998 cruise, Δ17O was generally <1‰ except for two samples, both collected south of the equator. There is thus evidence of S(IV) + O3 as an important source of sulfate under some conditions but not in others. Association with high Na+ concentrations (>4 μg m−3) is consistent with this source taking place in sea-salt aerosols.
4. GEOS-CHEM Model
 We use the GEOS-CHEM global three-dimensional (3-D) model of coupled aerosol-oxidant chemistry [Park et al., 2004] to simulate the Δ17O observations from the two cruises. The model (version 5.04) (available at http://www-as.harvard.edu/chemistry/trop/geos/geos_versions.html) uses assimilated meteorological data from the NASA Goddard Earth Observing System (GEOS) including winds, convective mass fluxes, mixed layer depths, temperature, precipitation, and surface properties. Meteorological data have 6-hour temporal resolution (3-hour for surface variables and mixing depths). Meteorological fields for 1996–1997 have a horizontal resolution of 2° latitude by 2.5° longitude (2° × 2.5°) with 46 sigma vertical levels (including 3 below 1 km for a column based at sea level). Meteorological fields for 1998 have 1° × 1° horizontal resolution with 48 sigma vertical levels (including 7 below 1 km for a column based at sea level). For input into GEOS-CHEM we degrade the horizontal resolution to 4° × 5°. We conduct two simulations, one for July 1996 to June 1997 and one for January to April 1998.
 The sulfur simulation is as described by Park et al.  with the addition of ship emissions of SO2 from the ship emissions assessment inventory [Corbett et al., 1999], sea-salt emissions from the surface ocean (two size classes partitioned at a radius of 1 μm), and sulfate formation through S(IV) + O3 in sea-salt aerosols. Sulfate produced by each oxidation pathway is transported as a separate “tracer” in the model with a corresponding Δ17O value. Total sulfate and Δ17O are determined locally by addition.
 Anthropogenic sources of sulfur in the model total 67.5 Tg S yr−1, including 4.2 Tg S yr−1 from ships. This anthropogenic sulfur is emitted as SO2 except for 3% as primary SO42− (5% in Europe) [Chin et al., 2000]. Natural sources include oceanic phytoplankton (12.4 Tg S yr−1 as DMS), volcanoes (5.5 Tg S yr−1 as SO2), and biomass burning (1.3 Tg S yr−1 as SO2). The emission flux of sea-salt (particle m−2 s−1) is calculated as a function of dry particle size and local 10 m wind speed following the empirical relationship of Monahan et al.  and is integrated over fine (0.1–1 μm) and coarse (1–10 μm) size ranges. All emissions are distributed vertically by mass in the local mixed layer.
 The simulation includes 10 chemical species transported in the model: DMS, SO2, sulfate (as six separate tracers), methanesulfonic acid, NH3, NH4+, NO3−, H2O2, submicron sea salt, and supermicron sea salt. Sulfate formation from oxidation of SO2 takes place by gas phase oxidation by OH and by in-cloud oxidation by H2O2 and O3. A cloud pH of 4.5 is assumed for the latter. The oxidation of S(IV) by O3 in sea-salt aerosols is described in section 5. We use monthly mean oxidant (OH, O3, and NO3) concentration fields, H2O2 production rates and photolysis frequencies, and total inorganic nitrate (HNO3 (g) plus aerosol NO3−) concentrations from a coupled aerosol-oxidant simulation as described by Park et al. .
 Wet deposition of aerosols is as described by Liu et al.  and includes contributions from scavenging in convective updrafts, rain out and washout from convective anvils and large-scale precipitation, and return to the atmosphere following reevaporation. Dry deposition velocities for sea-salt aerosols (and sulfate formed in sea-salt aerosols) are computed with the size-dependent scheme of Zhang et al.  integrated over each model size bin and accounting for hygroscopic growth as a function of relative humidity [Gerber, 1985]. Dry deposition velocities for all other species are computed with a standard resistance-in-series scheme based on Wesely  as described by Wang et al. .
 Our global sea-salt aerosol budget is given in Table 1, and global maps of surface (0–100 m altitude) air concentrations for January 1997 (pre-INDOEX 1997) and March 1998 (INDOEX 1998) are shown in Figure 2. Our sea-salt emission (5370 Tg yr−1) and burden (5.9 Tg) are within the range reported in the literature (Table 1). Differences between models reflect, in part, the accounting of different size ranges. Our simulated monthly average sea-salt concentrations over the Indian Ocean (Figure 2) show a similar spatial distribution to the [Na+] measurements from the two INDOEX cruises (Figure 3), with higher concentrations south of the equator due to stronger winds.
Table 1. Global Sea-Salt Aerosol Budgets Reported in the Literaturea
5. Sea-Salt Aerosol Alkalinity and Sulfate Production
 Sea-salt aerosol alkalinity is consumed rapidly by uptake of SO2 (g), H2SO4 (g), and HNO3 (g). To describe this process in the model, we start by calculating the fluxes of the three acids to sea-salt aerosols in the MBL column by assuming that the alkalinity is present in excess. We then compare the sum of these fluxes to the local emission flux of alkalinity. If alkalinity is indeed in excess, then the rate of S(IV) + O3 in sea-salt aerosols is calculated as the SO2 flux. We assume that the excess alkalinity is eventually removed by uptake of carboxylic acids [Mochida et al., 2003; Legrand et al., 2004] or hydrochloric acid (HCl) or by deposition to the oceans; it is not transported outside of the MBL in the model. If alkalinity is not in excess, we assume that SO2 (g), H2SO4 (g), and HNO3 (g) compete for the available alkalinity on the basis of their relative fluxes. Aerosol neutralization by NH3 is not significant above pH values of ∼6 [Jacob et al., 1986; Shi et al., 1999] and so is not considered here for sea-salt aerosols.
 Uptake of SO2 (g) by sea-salt aerosols is driven by the SO32− + O3 aqueous phase reaction with a rate constant k = 1.5 ± 0.6 × 109 mol L−1 s−1 [Hoffmann and Calvert, 1985]. For a pH = 8, a typical O3 concentration of 30 ppb, and a Henry's law constant for O3 of 1.1 × 10−2 mol L−1 atm−1 [Seinfeld and Pandis, 1998] we obtain a first-order loss rate constant for S(IV) in the aqueous phase kchem = k[O3][SO32−]/[S(IV)] = 4.4 × 10−1 s−1. This loss rate constant is sufficiently fast that the overall reaction rate is limited by SO2 transfer from the gas to the aerosol phase [Jacob, 2000; Liao et al., 2004]. Ozone, present at much higher concentrations than SO2, is in excess. We express the SO2 transfer rate constant (kt) following the formulation of Schwartz :
integrated over each particle radius bin [r0, r0 + Δr]. Here N(r)dr is the number concentration of sea-salt particles of radius [r, r + dr] derived from the Monahan et al.  emission formulation and allowing for hygroscopic growth as a function of local relative humidity, D = 0.2 cm2 s−1 is the gas phase molecular diffusion coefficient of SO2 in air, v = 3 × 104 cm s−1 is the mean molecular speed, and α = 0.11 is the mass accommodation coefficient for SO2 on liquid water surfaces [Worsnop et al., 1989].
 Uptakes of H2SO4 (g) and HNO3 (g) by sea-salt aerosols are similarly limited by transfer to the aerosol surface [Abbatt and Waschewsky, 1998; Song and Carmichael, 1999; Guimbaud et al., 2002]. We estimate the uptake rate of H2SO4 (g) by aerosols from the production rate of H2SO4 (g) through (R1), partitioned between sea-salt and H2SO4-HNO3-NH3-H2O aerosols on the basis of their locally computed relative surface areas in the model. For the latter we assume a lognormal size distribution with mean radius = 0.082 μm and geometric standard deviation of 2.0 to determine surface area. The sea-salt aerosol surface area is calculated by integrating the emission flux (particle m−2 s−1) for each size bin [Monahan et al., 1986]. The computation of HNO3 (g) concentrations in the aerosol-oxidant GEOS-CHEM simulation of Park et al.  is biased high because uptake by aerosols is considered only through the HNO3-H2SO4-NH3 system and does not include uptake by sea salt. To deal with this problem, we use the locally computed HNO3 (g) deposition flux to the ocean in GEOS-CHEM as representing, in fact, uptake by sea-salt aerosols. The lifetime of HNO3 (g) against deposition to the ocean surface (∼1 day) is sufficiently long compared with uptake by sea-salt aerosols (∼1 hour) that the latter will indeed dominate as long as the aerosols remain alkaline.
 We find in this manner that sea-salt aerosol alkalinity is usually titrated by excess acidity, on a timescale of <1 hour following emission, for most oceanic conditions. Exceptions are remote oceanic regions with high sea-salt emissions, such as the Southern Ocean. Uptakes of SO2 (g) and HNO3 (g) are both important for titrating sea-salt aerosol alkalinity, their relative importance depending on region (Figure 4). Uptake of H2SO4 (g) is negligible in comparison because of low surface concentrations and competition with condensation on H2SO4-HNO3-NH3-H2O aerosols. Titration by SO2 is most important at extratropical northern latitudes where anthropogenic emissions (including ship emissions) are high. Titration by HNO3 (g) is most important in the tropics where NOx emissions from soils, biomass burning, and lightning dominate over SO2 emissions.
6. Model Evaluation With INDOEX Measurements
 We ran two GEOS-CHEM simulations for the period of each cruise, one with S(IV) oxidation in sea-salt aerosols as described in section 5 and one without. The model was sampled as 24-hour averages along the track of each cruise. Since the model does not resolve the sulfate size distribution, the observed Δ17O data for the coarse and fine size ranges were combined to give a bulk Δ17O value for each sample (Δ17Obulk = χΔ17Ofine + (1 − χ)Δ17Ocoarse, where χ is the fraction of total nssSO42− in the fine size range). The comparison between simulated and observed bulk Δ17O is shown in Figure 5 for the pre-INDOEX 1997 (Figure 5, top) and 1998 INDOEX (Figure 5, bottom) cruises. Δ17O values in the simulation with no sea-salt chemistry are always <1‰, while the simulation with sea-salt chemistry shows a similar range as the measurements (∼0.5–3‰ for pre-INDOEX 1997 and ∼0.5–2‰ for INDOEX 1998) and a similar latitudinal structure (Δ17O > 1‰ only south of the equator). The latitudinal structure of Δ17O in the model is driven by sea-salt emissions, as illustrated in Figure 2 by the strong gradient of sea-salt concentrations along the cruise tracks.
 For the pre-INDOEX 1997 cruise we ran two additional sensitivity simulations. In the first sensitivity simulation we investigated the effect of an additional biogenic alkalinity source in sea-salt aerosols by doubling the alkalinity flux (0.14 eq kg−1) on the basis of the suggested 1–2.5 times enhancement of alkalinity indicated by calcium measurements from the western Pacific Ocean [Sievering et al., 2004]. This results in a 50–100% increase in the amount of sulfate formed in sea-salt aerosols (3–38% of total sulfate along the cruise track, compared to 2–24% in the standard simulation), with an increase in Δ17O sulfate values of 0.1–1.1‰ (Figure 6, top). This simulation reproduces the latitudinal trend in Δ17O values; however, unlike the observations, all Δ17O values are ≥1‰, and the maximum Δ17O value overestimates the observations by 0.8‰. Chlorophyll images from Sea-viewing Wide Field-of-view Sensor (SeaWIFS) (available at http://seawifs.gsfc.nasa.gov/seawifs.html) indicate that oceanic productivity is relatively low in the Indian Ocean during January to March 1998 (data are not available for January 1997). The contribution of biogenic alkalinity and its effects on the sulfur budget will vary with oceanic productivity [Sievering et al., 2004]. The impact of a biogenic alkalinity source would be better tested with sulfate Δ17O measurements in a region of high oceanic productivity and high acid flux such as the North Atlantic.
 In the second sensitivity simulation we assumed that OH uptake by sea-salt aerosols always maintains an excess of alkalinity [Laskin et al., 2003]. The Δ17O values in this simulation greatly overestimate the measurements (Figure 6). We conclude that regeneration of alkalinity does not, in fact, take place, consistent with the observations of Keene et al. .
7. Implications for the Sulfur Budget
Table 2 presents the GEOS-CHEM global budgets for sulfate produced by different oxidation pathways. For a general discussion of the GEOS-CHEM global sulfate budget, including comparisons to other models, see Park et al.  and Martin et al. . Sulfate formed by gas phase oxidation has the longest lifetime (6.1 days) because of a relatively high contribution from dry regions and high altitudes. The shortest lifetime (0.5 days) is for sulfate formed on coarse sea-salt aerosols. Wet processes account for 85% of deposition for total sulfate and dominate except for sulfate formed in coarse sea-salt aerosols. Sulfate formed on sea-salt aerosols accounts for 9% of global sulfate production but only 1% of the global sulfate burden. Liao et al.  found a much larger contribution to the burden (13%) because they assumed, following Laskin et al. , that reaction of OH with sea-salt aerosols would sustain alkalinity. As we have seen, this is not consistent with the oxygen isotope data.
Table 2. GEOS-CHEM Global Budgets for Sulfate Produced by Different Oxidation Pathways
As percentages of total deposition, which balances production.
SO2 + OH (gas phase)
S(IV) + H2O2 (in cloud)
S(IV) + O3 (in cloud)
S(IV) + O3 (fine sea salt)
S(IV) + O3 (coarse sea salt)
 In our model, sulfate produced via S(IV) + O3 in clouds is 4 times more abundant globally than sulfate produced by the same process in sea-salt aerosols. This is despite our assumption of a uniform cloud water pH of 4.5, greatly limiting the importance of the S(IV) + O3 relative to the S(IV) + H2O2 pathway. The contribution from the S(IV) + O3 in-cloud reaction is largely confined to the Northern Hemispheric winter when H2O2 is titrated [Kleinman and Daum, 1991; Feichter et al., 1996], but there are large uncertainties with this result involving SO2 oxidation on ice (in our model, in-cloud oxidation shuts down at temperatures below 258 K) and the competition from other nonphotochemical oxidants [Feichter et al., 1996; Kasibhatla et al., 1997; Leriche et al., 2003]. Δ17O measurements of sulfate in that environment could effectively resolve these issues.
Figure 7 shows the relative effect of sea-salt chemistry on simulated yearly mean SO2 and sulfate concentrations in the MBL. Concentrations of SO2 decrease by 10–30% over most of the oceans, up to a maximum of >70% over the Southern Ocean. Formation of sulfate in sea-salt aerosols could either increase total sulfate (by competing with SO2 deposition) or decrease it (because of faster dry deposition leading to shorter sulfate lifetime). We find that sulfate concentrations increase slightly in much of the MBL, up to a maximum of 30% in the tropical South Pacific. In that region, rapid wet deposition (due to strong convection) makes the negative effect on sulfate concentrations relatively less important. Using a global model, Gong and Barrie  previously examined the effect of sea-salt aerosols on the sulfur budget in the MBL due to condensation and coagulation processes and found decreases of 5–75% in sulfate concentrations in the MBL as a result of condensation of H2SO4 (g) followed by deposition. Our simulation does not account for this effect but indicates that it would be partially compensated by SO2 oxidation in sea-salt aerosols, not taken into account by Gong and Barrie .
Figure 7 also shows the impact of sea-salt chemistry on simulated yearly mean H2SO4 (g) production rates by (R1) in the MBL. The decrease is usually <30 but >50% over the Southern Ocean. This would tend to decrease the potential for formation of new particles over the oceans. The effect tends to be greatest where sea-salt aerosol concentrations are highest, and these are not regions propitious for formation of new particles in any case. Nevertheless, the widespread nature of the effect (10–30% over most of the oceans) suggests that it could be of some importance for model simulations of aerosol nucleation.
 Measurement of the mass-independent composition of the oxygen isotopes (Δ17O ≠ 0) of atmospheric sulfate aerosol provides quantitative insight into the oxidation pathways responsible for sulfate formation. A positive Δ17O implies aqueous phase formation involving oxidation of S(IV) by H2O2 or O3. A Δ17O > 1‰ specifically implies aqueous phase oxidation by O3 [Savarino et al., 2000]. The latter reaction can proceed at a significant rate only at relatively high pH (>6). Measurements presented here from two INDOEX ship cruises over the Indian Ocean show frequent occurrences of Δ17O > 1‰ in regions of high sea-salt concentrations but not elsewhere. This provides evidence for sulfate formation in alkaline sea-salt aerosols by the S(IV) + O3 pathway.
 We used the GEOS-CHEM global chemical transport model to interpret the Δ17O observations and to place them in a global context. We showed that the extent of sulfate formation in sea-salt aerosols is usually limited by the alkalinity flux associated with sea-salt aerosol emission and by competition for this alkalinity supply from HNO3 (g). Once the alkalinity of the sea-salt aerosols has been titrated, the pH drops abruptly, and sulfate formation shuts off. In the extratropical Southern Hemisphere where the sea-salt alkalinity flux often exceeds the availability of strong acids, sulfate formation in sea-salt aerosols is limited by the gas-to-particle transfer rate of SO2.
 We find that a model simulation including S(IV) + O3 oxidation in alkaline sea-salt aerosols, and accounting for acid titration of sea-salt alkalinity, can reproduce the range and latitudinal structure of the observed Δ17O sulfate values from the two INDOEX ship cruises. The latitudinal structure of Δ17O in the model is driven by sea-salt emissions. A sensitivity simulation with a doubled alkalinity flux, to account for a possible biogenic alkalinity (CaCO3) source, produces similar results in terms of the latitudinal trend in the Δ17O observations, with a 0.8‰ overestimate of the maximum Δ17O observation. The potential impact of a biogenic alkalinity source on the sulfur budget in the MBL could be better tested with sulfate Δ17O measurements in a region of high oceanic productivity and high acid flux such as the North Atlantic. Another sensitivity simulation not accounting for titration of alkalinity (i.e., assuming that alkalinity always remains in excess) greatly overestimates (by 3.8‰) the Δ17O observations. This argues against the hypothesis that sea-salt alkalinity is regenerated in the atmosphere by the NaCl + OH reaction [Laskin et al., 2003].
 We applied the GEOS-CHEM model with the above constraints to quantify the impact on the atmospheric sulfur budget of sulfate formation in sea-salt aerosols. Sulfate formed in sea-salt aerosols has a short lifetime (0.5–1.7 days) compared to sulfate formed in the gas phase or in clouds (3.1–6.1 days) and accounts for 9 and 1% of total global sulfate production and burden, respectively. This sea-salt chemistry decreases SO2 concentrations in the MBL by typically 10–30% but by >70% over the Southern Ocean. It increases sulfate concentrations by up to 30%. This chemical enhancement of sulfate in the MBL offsets, at least in part, the depletion due to condensation of H2SO4 (g) onto sea-salt particles followed by dry deposition [Gong et al., 2002]. Sulfate formation in sea-salt aerosols also leads to reductions in the rate of gas phase H2SO4 formation by 10–30% over much of the world's oceans. This would tend to decrease the potential formation of new particles over the oceans and hence any possible thermostat involving production of CCN from temperature-dependent DMS emissions.
 Δ17O measurements of sulfate aerosols collected at various continental sites also show values >1‰ [Lee et al., 2001; Lee and Thiemens, 2001; Bao and Reheis, 2003], indicating potential sources of continental alkalinity. Laboratory experiments have observed sulfate formation on dust when exposed to SO2 and O3 [Ullerstam et al., 2002], perhaps at least partially explaining these large continental Δ17O values. Ice core sulfate Δ17O measurements are available from Greenland and Antarctica extending through the last interglacial period and show values in the range of 0.2–4.8‰, varying with climate on the glacial-interglacial timescale [Alexander et al., 2002, 2004]. Interpretation of these ice core data with a global 3-D paleoclimate model could provide important new information on trends in sulfur and oxidant budgets over Earth's history.
 Financial support for this project was provided by a NOAA CGC postdoctoral fellowship to B. Alexander and by the NSF Atmospheric Chemistry Program. The GEOS-CHEM model is managed at Harvard University with support from the NASA Atmospheric Chemistry and Modeling Analysis Program. B. Alexander is a Daly postdoctoral fellow in the Department of Earth and Planetary Sciences, Harvard University.