The rainwater concentration of major components is given in Table 1 and of trace metals in Table 2. Ship and decimal day number (DOY) identify the rain events. Samples represent the whole duration of each rain event as encountered on the ship except for one case (87.5R) when two fractions were collected. For comparison, we have selected some results from rain chemistry measurements in India which we consider to be representative for rural conditions. These are median concentrations from station Pune (in the western part of India) during days with trajectories indicating transport from the Bay of Bengal across India to Pune [Norman et al., 2001] and a yearly weighted mean from station Bhubaneswar in eastern India [Granat et al., 2001].
3.1. Spatial Distribution of pH, nss-SO42−, NH4+, and nss-Ca2+
 As shown in Figure 2a the variation in pH was modest, ranging from 4.8 to 5.4. This narrow range of pH was in sharp contrast to most measurements over the Indian subcontinent, which in general showed much higher pH values [Rao, 1997]. Many of the measurements summarized by Rao were made using bulk collectors - often in urban environments - and are likely to have been affected by local sources of alkaline dust. pH values in better agreement with the ones observed during this study have been reported by two recent studies, one outside Bhubaneswar in eastern India [Granat et al., 2001] and one from Pune in western India during easterly winds [Norman et al., 2001]. In these two studies great care was taken to obtain values representative for larger regions.
 The acidity of precipitation is dependent on a balance between the uptake of acid and alkaline compounds. The concentration of H+ in the raindrop equals the difference between the sum of the negative ions and all other positive ions. In India, as in many other continental regions, the dominating negative ions are nss-SO42−, NO3−, and HCO3− and the dominating positive ions, in addition to H+, are nss-Ca2+ and NH4+. For the range of pH encountered over the ocean the contribution of HCO3− was neglected. NO3− did contribute to the balance over the ocean, but its importance was less than that of SO42−. As a consequence, we have focused our discussion mainly on the remaining three ions, i.e., nss-Ca2+, NH4+, and nss-SO42−, assuming that the sea salt components do not contribute to the balance. The contribution from acetate and formate was usually relatively small, with exception of the “A” cases (air from Arabian Peninsula) and case 58.9R, see Table 1.
 The association between the trajectories and the concentrations of nss-Ca2+, NH4+, and nss-SO42− is shown in Figure 2b. No obvious spatial pattern was evident in the concentrations of these ions, but they vary systematically with trajectory type, c.f. Table 1. For example, the average concentration of NH4+ during days with trajectories originating from NH and SH was 3.3 and 0.4 μmol dm−3, respectively. The corresponding difference for nss-Ca2+ and nss-SO42− was much less. The influence of trajectory types will be discussed in more detail below.
3.2. Sources and Transport
 The sources of the ionic components and metals observed in the rainwater samples can be either terrestrial or marine. Terrestrial sources include fossil fuel combustion and other industrial activities, biomass burning and other agricultural activities, and wind-blown soil-dust. The Indian subcontinent can be suspected to be the most important origin in these cases, but more distant sources can not be excluded. Of the analyzed components the most important marine sources are primary sea salt particles from sea spray and secondary aerosols produced from gaseous precursors like dimethyl sulfide (DMS). Great care was taken to avoid pollution from the ships themselves, cf. section 2.1.
 In order to interpret the variability of the observed concentrations of the various constituents, as exemplified in Figures 2a and 2b, the statistical correlation between their concentrations and their relation to trajectories were investigated. A separate discussion of the sea salt source is also included in this section.
3.2.1. Correlation Between Components
 A factor analysis (principal component analysis with Varimax rotation) including all analyzed rain components gave three factors with the following components (see Table 3 for values of multiple correlation coefficients, as r): (1) NH4+, nss-K+, nss-Ca2+, NO3−, and nss-SO42−, (2) acetate and formate, and (3) MSA.
Table 3. Correlation Coefficients for Ionic Components in Rain
 This statistical analysis alone could not distinguish between soil derived components, such as nss-Ca2+ and those derived from gaseous precursors, such as nss-SO42− and NO3−, implying that their source regions are similar.
 MSA did not correlate with any component including nonsea-salt fractions, with the exception of Na+ where a modest correlation was found (0.72). The correlation with Na+ may be due to a larger aerosol surface area enhancing the transfer from vapor phase as previously suggested [e.g., Quinn et al., 1993; Huebert et al., 1993].
 Acetate and formate were well correlated with each other (0.92). Molar concentration of formate was about 1.4 times that of acetate. Both components seem to have derived from the same source located in the Northern Hemisphere (NH) but not related to sources of the other compounds. This could be explained by a gradual production from long-lived gaseous precursors, which can be expected to be more abundant in NH than in SH.
 Nss-F− was lower in SHo samples but did not correlate well with any other compound. Its presumed origin in combustion and process industry did not show up as a correlation with nss-SO42−.
 Among the trace metals two groups could be distinguished (not shown here). One consisting of Fe and Al, well correlated with each other and with nss-Ca2+, probably representing crustal material. The other consisted of V, Cd, Tl, and Pb that were well correlated with each other (r > 0.9 uncertain due to few samples) and with nss-SO42−, thus indicating anthropogenic sources.
3.2.2. The Sea Salt Source
 The production of sea salt aerosols due to bursting of bubbles in seawater is known to be related with wind speed. Two different size ranges of particles have been identified, film drops and jet drops. Day and Lease  found the film drops, released into the air when the bubble film of water burst, to be around 100 nm in diameter while Blanchard and Woodcock  found jet drops that were detached from the up moving water replaced the bubble film around 1 μm in diameter tailing toward larger sizes with increasing wind speeds. Several studies have found the logarithm of the aerosol number or mass concentration to be linearly related to wind speed. We found a similar relation to hold also for the concentration of sea-salt in precipitation:
 The best correlation (r2 = 0.69) was found if the maximum 24 hours average wind speed during the last 3 days backwards along the trajectories was used instead of the local wind speed. This indicates that the Na+ in the rain samples was not only of local origin. This studies rainwater slope of 0.11 was in good agreement with those reported for both number and sodium mass over the central Arctic Ocean [Nilsson et al., 2001, [Leck et al., 2002] and applied to the jet drop mode sized particles.
 The relative concentration of the sea salt components Cl− and Br− in rain may be modified from that of seawater by atmospheric processes or by contributions from terrestrial sources. Chlorine can be lost from the aerosol as gaseous HCl when sulfate, produced by aqueous oxidation of SO2 or condensation of sulfuric and methane sulfonic acids, displaces chlorine [Keene et al., 1998].
 The molar ratio for a specific particle of pure seawater NaCl introduced into the atmosphere is 1.16. The ratio at a later time will depend on how long it has been in the atmosphere, its original mass, and the rate at which it accumulates acid. Table 4 shows the mean of Cl−/Na+ molar ratios encountered as a function of aerosol size (fpm and cpm). The mean fpm ratios as very different from the cpm ratio suggesting that the submicron particles were in general more likely to have an acid component than the supermicron particles. This in agreement with reported results elsewhere (e.g., Leck et al., submitted manuscript, 2000). The mean cpm Cl−/Na+ ratio was slightly lower than that of pure seawater, which suggests the importance of a local or near local NaCl aerosol source.
Table 4. Mean Aerosol Concentration for Fine Particle Mass (fpm) and Coarse Particle Mass (cpm) During the Rain Eventsa
|Aerosol concentrations, nmol m−3||Cl−/Na+||Na+||NH4+||K+||nss-K+||Mg2+||Ca2+||nss-Ca2+||MSA||Cl−||NO3−||SO42−||nss-SO42−|
|total (fpm + cpm)||1.03||90.0||16.6||3.9||2.0||9.9||3.5||1.4||0.3||92.3||9.4||19.2||13.8|
|% fpm of total|| ||2||100||49||95||2||11||14||67||1||1||28||99|
 The nss-Cl− concentrations in rainwater from Ronald H. Brown were not significantly different from zero except for a slight deficit (5% excess of cations) in one event (79.3R). A similar deficit was then also obtained for Mg2+ and Ca2+ indicating an “extra” contribution of Na+ rather than a loss of Cl−. On Sagar Kanya an excess nss-Cl− was found in event (54.3S) and a deficit in event SK1 (28.7S). No simple explanation could be given in these cases and it was concluded that there was no overall indication of any loss of Cl− in precipitation compared to seawater composition. The result is further consistent with a wind speed generated jet drop source of rainwater sea-salt.
 Nss-Br− concentration can be decreased in the aerosol as a result of bromine oxidation followed by escape of volatile bromine compounds. The rain chemistry data showed that any deficit or excess Br− was less than 10%, which is the limit of analytical accuracy due to interference from varying concentrations of NO3−.
3.2.3. Relation to Air Trajectories
 Trajectories where the air parcel had been exposed to Indian sources (c Ind) showed the highest concentration of NH4+, nss-K+, nss-Ca2+, NO3−, and nss-SO42−, cf. Table 1. For nss-F−, formate, acetate, and MSA other events had similar or higher concentrations. After a transport of three to four days over sea the concentrations of NH4+, NO3−, and nss-SO42− were 0.4 to 0.5 of those over land if we choose the data from Pune (west) and Bhubaneswar (east) as reference. For nss-Ca2+ the ratio was somewhat lower, about 0.3 of this land-based reference. The faster depletion of nss-Ca2+ from land to sea was in agreement with its expected distribution on larger particles.
 Many rain chemistry measurements in India have given nss-Ca2+ concentrations up to several times higher than our reference. Compared to those the nss-Ca2+ concentrations over sea reported here were very low indeed. This emphasized the possibility of local influence of many continental measurements, especially of soil dust components.
 The molar ratios of NH4+ and NO3− to nss-SO42− were 1.3 and 1.0 respectively while we calculate the ratios of emitted precursors (NH3 and NOx to SO2) in India to be 5.7 and 1.1 respectively. The emission estimates were taken from Zhao and Wang , van Aardenne et al. , and Streets et al. , respectively, and adjusted to 1997. The high emission of NH3 compared to SO2 and NOx is noteworthy. NH3 was obviously depleted more rapidly than the other two components. Data from rural India on formate and acetate concentrations in rain during winter do not seem to be available. Compared to measurements in a semiurban area (Dayalbagh, near Agra, [Kumar et al., 1996]) our data were a factor 10 lower.
 The nss-K+ to nss-SO42− molar ratio in rain in air of NH origin was 0.10. This can be compared with molar ratios of K+ to SO2: (1) from biomass burning ranging from 0.3 to 1.6 depending on type of fire, with the lowest ratio for biofuel [Andreae and Merlet, 2001], (2) with 0.05 from fossil fuel combustion, and (3) with 0.07 in rain in Northern Europe with almost no biomass burning. This suggests that there was only a small, less than 15%, contribution to nss-K+ and nss-SO42− from biomass burning. This is in line with the results from the more comprehensive aerosol measurements during INDOEX, from which Mayol-Bracero et al.  estimated that fossil fuel contribution to nss-SO42− in aerosol over the Indian Ocean is 60–90%. Transport from Arabia and some surrounding areas followed by about seven days over the sea gave generally low concentrations of the nss components. Since the emissions of NOx, SO2, and probably also NH3 are much lower, and of Ca much higher, in this area compared to India [Benkovitz et al., 1996; Tegen and Fung, 1994], one would expect a corresponding influence on the rainwater concentrations. However, the concentration of nss-SO42− in trajectory group A, 7d was only a factor two to three lower than in air from India.
 Two of the samples from the SHo had the lowest concentrations of all events for nss components except MSA. Concentrations in the third SHo sample (37.9S) were higher, comparable to those from the NHo with regard to nss-SO42− and nss-Ca2+. The SHo samples were characterized by low values of the following molar ratios: NH4+/nss-SO42− (0.1), nss-K+/nss-SO42− (<0.05), and NO3−/nss-SO42− (0.1). The nss-Ca2+/nss-SO42− ratio (0.14) was low but not much lower than in other trajectory groups.
 The contribution of NO3− to acidity (molar ratios) ranged from one third to about equal of nss-SO42− in NHo but was insignificant in SHo. NH4+ was a factor of two to three more important than nss-Ca2+ for neutralizing the acids. The sum of acetate and formate was similarly important as NO3− for the acidity in the SHo samples. The contribution from MSA itself to the ion balance was insignificant but the formation of SO42− from DMS may have been important in air with SHo.
3.3. Aerosol Scavenging Ratios
 The simultaneous observations of the chemical composition of rainwater and size resolved aerosols provide a unique opportunity to draw conclusions about the aerosol scavenging processes. This has been done by considering the aerosol scavenging ratio (SR) defined as [Galloway et al., 1993]:
where Crain equals the concentration of a component in rain (nmol kg−1), Caerosol its concentration in air (nmol m−3), and Da the density of air. SR is thus expressed in kg of air (kg of water)−1. Table 5 shows SR calculated for aerosol fpm and cpm separately. Fpm SRs were calculated for components dominating the submicron fraction (Table 4), that is NH4+, nss-K+, and nss-SO42−. Cpm SRs were first calculated for Na+, Mg2+, and Cl− which had their mass almost entirely in the supermicron fraction. In the case of nss-Ca2+ and MSA, which were distributed over both size fractions, the rainwater concentrations were subtracted with the contribution from fpm based on the SR for nss-SO42− before calculation of SRcpm:
Table 5. Scavenging Ratios (SR) for Fine Particle Mass (fpm), Coarse Particle Mass (cpm), and Total Particle Mass (total) from Ronald H. Browna
| ||Precipitate amount, mm||Precipitate intensity, mm h−1||fpm||fpm||fpm||total||cpm||cpm||cpm||cpm||cpm||cpm|
|NH4+ SR||nss-K+ SR||nss-SO42− SR||MSA SR||NO3− SR||Na+ SR||Mg2+ SR||Cl− SR||nss-Ca2+ SR||MSA SR|
|54.9R||15.4||20.0||427|| || ||168||106||694||656||758||169||91|
|78.9R||0.4|| ||256||278||441||424||670||776||767||835|| ||510|
|81.5R||1.0|| ||279||170||236||273||274||1055||1063||1560||332|| |
 If both the submicron and supermicron aerosol were assumed to be released from the atmospheric boundary layer and processed by the clouds with the same efficiency, the below-cloud scavenging of cpm could be taken as the difference between SRcpm and SRfpm. This assumption neglects possible enrichment of soil dust particles due to ice phase processes as reported by Puxbaum and Tscherwenka  for colder regions.
 A wide variability in SRfpm was observed with a tendency for lower SRs with higher rainfall intensity, as shown in Table 5. The SRs decreased with rainfall intensity approximately according to:
with SR in kg kg−1 and i, the rainfall intensity, in mm h−1. The variation on event basis was similar for nss-K+ and nss-SO42− while that of NH4+ showed a more narrow range. The median SRfpm (Table 5) was very similar for the three components NH4+, nss-K+, and nss-SO42− (∼250 kg kg−1).
 The median SRcpm (∼750 kg kg−1) for Na+, Cl−, and Mg2+ varied between events in a similar fashion as the submicron aerosol components, with a tendency for the SRs to decrease with rainfall intensity approximately as:
 The SRtotal for nss-Ca2+ (distributed mainly in cpm) gave a median of 330 kg kg−1, lower than that for the sea salt and with event-to-event variations much better correlated with the SRfpm for nss-SO42− than with the SRcpm for Na+.
 In the case of MSA the SRtotal was even closer to the nss-SO42− SRfpm median value. This is probably because MSA, although appearing in both fpm and cpm, was mainly distributed in the submicron size fraction.
 The SRcpm for NO3− was substantially lower (median 335 kg kg−1) than for the cpm components of sea salt origin. In view of a possible contribution from gaseous HNO3 a higher value might have been expected. There was thus a clear indication that nitrates are removed primarily by in-cloud scavenging, as in the case of nss-SO42−. The reason for the inefficient removal below cloud could be that the nitrate was distributed toward the smaller end of the supermicron size fraction compared to the other sea salt components.
 Both the calculated SRcpm subcloud and in-cloud scavenging and the SRcpm for below-cloud scavenging alone (taken as the difference between SRcpm and SRfpm) were within the estimated range given by Seinfeld and Pandis . Nevertheless, the larger SRcpm calculated for the supermicron sea salt (Na+, Mg2+, and Cl−) components relative to the major submicron components indicates that the sea salt aerosols were removed by both in-cloud and subcloud scavenging.
 On one occasion the rain was divided into sequential periods (87.4R in Table 1 represents the total rain and 87.5R the last 9 of the total 41 mm). Concentrations at the end were generally lower but the remaining fraction differed between the components, about 0.3 for sea salt, NH4+, and NO3− but larger, 0.7, for nss-SO42−. For nss-K+ and nss-Ca2+ the result was scattered (0.0 and 0.5) but the uncertainty in the calculation is large for these ions.