3.1. Air Mass Origins and Characteristics
 Air mass back trajectories for a number of samples are plotted in Figure 1. The beginning of the cruise was characterized by air masses that originate from the Indian subcontinent. The different air circulation patterns observed during this period are depicted in Figures 1a, 1b, and 1c, which are representative of the group of samples from IO97-01 to IO97-10. Trajectories in sample IO97-11 (not shown) appear to be transitional between the first and the second group of samples, the latter of which includes samples IO97-12 through IO97-27. The distinguishing characteristic in trajectory pattern between the two groups is the main source of the air masses. As observed in Figures 1d, 1e, 1f, 1g, and 1h, the air masses were swept around in a clockwise manner from the Saudi Arabian peninsula, thus most likely transporting mineral dust from the arid regions of Saudi Arabia to the middle of the Arabian Sea from an apparently northeasterly direction at the collection point. Thus the wind direction observed on board the ship would have erroneously suggested air mass origins in India and Pakistan. Since air mass origins during this second group of samples are similar to those expected during the inter-monsoon [Johansen et al., 1999; Siefert et al., 1999], it is appropriate to report average data from both groups of samples when discussing distinctions between the different monsoon seasons.
 Within the second group of samples, air masses vary considerably in their relative transport speeds. Samples IO97-11 through IO97-15 and IO97-20 through IO97-26 (Figures 1d, 1f, and 1g) are representative of air masses that have traveled from the Saudi Arabian peninsula over considerable distances in a relatively short period of time and are, therefore, expected to carry relatively more crustal material compared to air masses sampled in samples IO97-16 through IO97-19 (Figure 1e) and sample IO97-27 (Figure 1h), during which wind speeds were lower.
3.2. Chemical Makeup of Aerosol
 Of the 33 elements detected by ICP-MS, the concentrations for elements As, Ru, Cd, Cs, and Sb were found to be below their detection limits and are, therefore, excluded from further discussion. Furthermore, principal component (PC) analyses [Johansen et al., 2000; Siefert et al., 1999] have shown high background concentrations for Cr, Ni, Cu, Mo, and Sn that seem to be attributable to artifacts introduced by the filter material and/or the acids used for the elemental extraction.
 Average values for the remaining 22 elements in the coarse and fine fractions for the two groups of atmospheric conditions are listed in Table 1. In order to examine the similarities among these elements and to investigate Fe(II), a PC analysis was performed on 11 of the trace metals in addition to Fe(II) and some anions and cations. The results are presented in Table 2. Four samples (IO97-08, IO97-18, IO97-19, and IO97-27) had to be excluded in this analysis because of miscellaneous data gaps in the anions and Fe(II) concentrations (vide supra).
Table 1. Average, Minimum, and Maximum Atmospheric Trace Metal Concentrations in Coarse and Fine Aerosol for Group 1 and Group 2 Samplesa
|Element||Group 1 (IO97_01-IO97_10)||Group 2 (IO97_11-IO97_27)|
|Average ± SD||Min.||Max.||Average ± SD||Min.||Max.|
|Na-coarse, μg m−3||0.39 ± 0.14||0.20||0.61||0.49 ± 0.24||0.14||0.88|
|Na-fine, μg m−3||0.21 ± 0.05||0.13||0.29||0.26 ± 0.13||0.09||0.61|
|Mg-coarse, μg m−3||0.13 ± 0.06||0.07||0.23||0.16 ± 0.09||0.06||0.38|
|Mg-fine, μg m−3||0.07 ± 0.02||0.05||0.11||0.10 ± 0.04||0.06||0.21|
|Al-coarse, μg m−3||0.32 ± 0.13||0.16||0.56||0.30 ± 0.24||0.08||0.93|
|Al-fine, μg m−3||0.15 ± 0.03||0.10||0.21||0.21 ± 0.09||0.11||0.04|
|K-coarse, μg m−3||0.05 ± 0.04||0.01||0.15||0.10 ± 0.09||0.02||0.35|
|K-fine, μg m−3||0.19 ± 0.06||0.09||0.30||0.15 ± 0.09||0.04||0.38|
|Ca-coarse, μg m−3||0.31 ± 0.17||0.13||0.60||0.38 ± 0.19||0.13||0.74|
|Ca-fine, μg m−3||0.12 ± 0.04||0.08||0.20||0.17 ± 0.9||0.09||0.39|
|Sc-coarse||0.083 ± 0.036||0.038||0.142||0.070 ± 0.058||0.022||0.230|
|Sc-fine||0.039 ± 0.007||0.027||0.053||0.047 ± 0.017||0.027||0.094|
|Ti-coarse||25.99 ± 10.90||13.57||45.14||24.02 ± 16.42||5.90||64.80|
|Ti-fine||11.85 ± 1.98||8.21||14.74||16.82 ± 6.83||10.04||32.39|
|V-coarse||0.73 ± 0.30||0.36||1.25||0.69 ± 0.48||0.18||1.92|
|V-fine||1.18 ± 0.16||0.86||1.38||1.19 ± 0.42||0.50||1.98|
|Mn-coarse||6.48 ± 2.57||3.70||10.37||5.09 ± 2.55||1.43||10.52|
|Mn-fine||4.69 ± 1.22||3.34||6.38||5.26 ± 2.28||2.29||9.40|
|Fe-coarse, μg m−3||0.39 ± 0.20||0.19||0.89||0.31 ± 0.17||0.08||0.70|
|Fe-fine, μg m−3||0.33 ± 0.13||0.17||0.53||0.41 ± 0.25||0.15||0.89|
|Zn-coarse||1.95 ± 0.44||1.29||2.63||0.99 ± 0.49||0.30||1.83|
|Zn-fine||9.72 ± 3.68||5.56||17.32||3.98 ± 2.15||1.55||9.16|
|Ga-coarse||0.14 ± 0.05||0.07||0.24||0.12 ± 0.07||0.03||0.28|
|Ga-fine||0.08 ± 0.01||0.06||0.10||0.09 ± 0.03||0.05||0.17|
|Ge-coarse||0.027 ± 0.008||0.027||0.039||0.021 ± 0.010||0.007||0.045|
|Ge-fine||0.045 ± 0.013||0.026||0.066||0.032 ± 0.010||0.017||0.050|
|Se-coarse||0.081 ± 0.030||0.036||0.137||0.069 ± 0.036||0.036||0.142|
|Se-fine||0.306 ± 0.094||0.214||0.514||0.257 ± 0.100||0.113||0.490|
|Ba-coarse||2.19 ± 0.81||1.12||3.64||1.64 ± 0.89||0.47||3.56|
|Ba-fine||1.01 ± 0.16||0.66||1.29||1.18 ± 0.42||0.62||2.27|
|La-coarse||0.17 ± 0.06||0.08||0.27||0.14 ± 0.10||0.04||0.40|
|La-fine||0.09 ± 0.1||0.06||0.11||0.11 ± 0.04||0.06||0.23|
|Ce-coarse||0.36 ± 0.14||0.18||0.59||0.33 ± 0.23||0.08||0.89|
|Ce-fine||0.18 ± 0.03||0.12||0.24||0.24 ± 0.10||0.13||0.48|
|Sm-coarse||0.031 ± 0.012||0.015||0.049||0.026 ± 0.017||0.008||0.067|
|Sm-fine||0.015 ± 0.003||0.009||0.019||0.018 ± 0.006||0.011||0.033|
|Eu-coarse||0.0078 ± 0.0028||0.0045||0.0123||0.0056 ± 0.0035||0.0018||0.0149|
|Eu-fine||0.0034 ± 0.0004||0.0028||0.0041||0.0043 ± 0.0013||0.0027||0.0069|
|Hf-coarse||0.020 ± 0.008||0.010||0.033||0.022 ± 0.010||0.010||0.039|
|Hf-fine||0.011 ± 0.003||0.008||0.016||0.021 ± 0.009||0.009||0.041|
|Pb-coarse||0.80 ± 0.18||0.52||1.09||0.39 ± 0.18||0.02||0.65|
|Pb-fine||6.78 ± 1.82||4.67||10.46||3.52 ± 1.70||1.32||6.88|
|Th-coarse||0.052 ± 0.020||0.023||0.088||0.047 ± 0.39||0.012||0.149|
|Th-fine||0.024 ± 0.004||0.015||0.032||0.031 ± 0.014||0.016||0.071|
Table 2. Varimax Rotated Principal Component Matrixa
|Element\Component||1 Crustal (24.6%)||2 Anthrop. (19.4%)||3 Crustal High in Ca and Mg (14.2%)||4 Sea Salt (11.3%)||5 Fe(II)-fine (6.2%)||6 MSA, NSS-SO42—coarse (5.8%)||7 Fe, Mn-fine (4.8%)||8 (3.9%)|
 The first eight components of this analysis carry eigenvalues larger than 1 and account for a cumulative variance of 90.3% of the data set. The variance described by each component appears across the top of the table.
 The first component, which is representative of a crustal source, has relatively “large” (i.e., values close to 1) values for typical crustal material tracers, such as Al and Sc (as well as Ti, Ba, La, Ce, Sm, Eu, and Th, not shown). No other known sources exist for these elements, therefore, they can be used reliably as indicators of crustal material. Some of the other elements (K, V, Mn, and Fe) that can be associated with gas-to-particle reactions of anthropogenic pollutants from combustion sources may still display crustal characteristics in the coarse fraction since gas-to-particle reactions result in fine particles. Concentrations for the crustal tracer Al are plotted as a function of sample ID in Figure 2a. For comparison, the factor scores for the crustal component (PC 1) are plotted in Figure 3a. As expected, the two variables trace each other closely.
Figure 2. Coarse and fine fraction trace metal concentrations versus sample ID. for (a) Al, (b) Pb, (c) Ca, (d) Na, and (e) Fe.
Download figure to PowerPoint
Figure 3. Factor scores for (a) the first three and (b) the second three principal components extracted in the principal component analysis in Table 2.
Download figure to PowerPoint
 The second component in the PC analysis reflects an anthropogenic source, typified by Pb and Zn in both size fractions (and Ge and Se, not shown) as well as K in the fine fraction. While Pb and Zn reach the atmosphere through burning of fossil fuel and smelting operations, respectively, K is a product of biomass burning and waste incinerators [Andreae, 1983; Echalar et al., 1995; Fishman et al., 1999]. Figure 2b presents the coarse and fine Pb concentrations in stacked bars as a function of sample ID. The plots for Pb, Zn, and K (latter two not shown) are all similar, and as expected for anthropogenic tracers, their concentrations are larger in the fine fraction as compared to the coarse fraction. The geometric mean for the relative abundance of Pb in the fine fraction is 89 ± 3%, for Zn it is 80 ± 8%, and for K it is 67 ± 16%. Factor scores for PC 2 in Figure 3a carry the same fingerprint as Pb. Consistent with our observations of the air mass back trajectories, PC 2 is predominant in samples of group 1, which are characteristic of the northeast monsoon. Lead and zinc concentrations increase by factors of 2.2 and 2.3, respectively, between samples of group 1 and 2.
 The third component (PC 3) represents another crustal source, rich in Ca and Mg. Upon inspection of its factor scores in Figure 3b, it is evident that this crustal component becomes of importance during the latter part of samples in the second group, especially in samples IO97-20 and IO97-24. Air mass back trajectories in Figure 1 indicate that these air masses were transported with relatively greater wind speeds from the Saudi Arabian Peninsula and as far as northeastern Africa. The apparently distinct chemical signatures between the two crustal components may also be a function of aerosol particle size and the different chemical transformations occurring during transport. Owing to the large correlation of PC 3 with the water soluble fractions of Ca and Mg (Ca2+ and Mg2+ in Table 2), it is likely that this crustal component carries a gypsum, calcite, and/or limestone fraction, all of which are typical constituents of clay minerals. A clay mineralogical study of the aerosol material over the Arabian Sea [Chester et al., 1985] revealed that illite is the dominant clay mineral. Illite was also observed to be the predominant clay mineral in most sediments from the northern Arabian Sea [Goldberg and Griffin, 1970].
 The fourth principal component (PC 4) is representative of the sea salt, and its factor scores are plotted in Figure 3b. Water-soluble Na (Na+, in Table 2) is used as the primary sea-salt tracer; however, Na, as plotted in Figure 2d, can potentially include a crustal source because it was determined by ICP-MS on a section of acid leached filter material. Within the analytical errors of our techniques, however, the crustal Na was negligible. The sea-salt component usually correlates well with the wind speed due to the increased wave action on the water surface, which leads to sea-salt aerosol formation; however, due to the relatively low wind speeds encountered during the present cruise, this correlation is only weak.
 PC 5 and PC 6 will be discussed below in conjunction with iron speciation. Fe and Mn in the fine fractions correlate with each other in PC 7. The source of fine Fe and Mn could arise from local incinerators or even from our research vessel's smoke stack. The same elements were observed to be enriched during a previous cruise on the R/V Meteor [Siefert et al., 1999]. Since both these vessels use diesel fuel and electric power generators to propel the ship, V was not expected to be seen in the ship's plume [Hopke, 1985]. However, Fe and Mn during the Meteor cruise were enriched in collected samples when the sampling sector system was malfunctioning. Thus there is a slight chance that although the sector sampling system was working well during the whole present cruise, minute amounts of the ship's plume may have reached the sampling setup during the ship's maneuvers. Fe concentrations are plotted in Figure 2e. The eighth component was retained in Table 2 but does not have any physical meaning other than that it seems to be slightly representative of the wind speed.
 As with the inter-monsoon samples collected in 1995 [Johansen et al., 1999], the crustal average according to Taylor and McLennan  appears to adequately represent the sampled mineral dust; however, an additional crustal Ca source seems present in all the samples. On the basis of the crustal average and the observed Al concentrations, the mineral dust concentrations averaged 5.7 ± 1.7 μg m−3 (67.8% in the coarse fraction) during the first group of samples and 6.1 ± 3.6 μg m−3 (58.4% in the coarse fraction) for the second group of samples. These two average crustal masses do not prove to be statistically different, thus indicating that the crustal component during the northeast monsoon and inter-monsoon may be of comparable magnitude. During the inter-monsoon of 1995, Johansen et al.  observed a very similar crustal average of 5.85 ± 4.24 μg m−3, while during the southwest monsoon, concentrations were considerably lower, 0.66 ± 0.43 μg m−3. Rhoads et al.  found very similar values of 6.2 ± 4.4 μg m−3 during the months of March and April of 1995. Chester et al.  reported mineral dust concentrations decreasing from 15–20 μg m−3 in the north to 0.01–0.25 μg m−3 below 35°S in the far Southern Ocean. Pease et al.  obtained mineral dust concentrations of 40 μg m−3 during the northeast monsoon of 1995, while Savoie et al.  measured dust concentrations of only 1.01 ± 0.81 μg m−3 during the northeast monsoon of 1979.
3.3. Fe and Fe(II)
 The PC analysis in Table 2 shows that coarse Fe seems to be of crustal origin (PC 1), while most of the fine Fe, which is comparable in magnitude to the mass in the coarse Fe, correlates with fine Mn (PC 7) for which the source is not obvious. On the basis of enrichment factor analysis (not shown), fine iron is on average 2.4 times the value expected from the crustal contribution, whereby there is no statistical difference between the group 1 and group 2 samples.
 Ferrous iron in the coarse fraction seems to be crustal derived (PC 1), while ferrous iron in the fine fraction (PC 5) does not correlate with any component but itself. The implications are that Fe(II) in the fine fraction must be regulated by either one or more unknown parameters in addition to a combination of the existing parameters. Assuming that the concentrations of all other pH modulating constituents in the aerosol particle remain unchanged, the biological NSS-SO42− contribution should lead to an acidification of the aerosol. However, from this study it is not clear that the biogenically derived NSS-SO42− acidity influences the concentration of Fe(II), as proposed by Zhuang et al.  and indicated by Johansen et al. .
 The three extracted Fe(II) portions for both coarse and fine fractions are presented in stacked bar plots in Figures 4a and 4b, respectively. Superimposed, in dashed lines, are the corresponding Fe concentrations determined by ICP-MS.
Figure 4. Fe(II) concentrations versus sample ID. (a) Fe(II) in the coarse fraction, (b) Fe(II) in the fine fraction, and (c) total Fe(II) released after 5 min in ferrozine (see text for details).
Download figure to PowerPoint
 Although, as mentioned earlier, there is an increased possibility of ferrozine reacting with Fe during the 22-hour extraction experiment, it is peculiar that the coarse and fine fractions exhibit widely different behaviors with regard to the amount of Fe(II) released in the three consecutive extraction steps. In the coarse fraction the Fe(II)22hrFZ (white columns in Figure 4a) comprised 80% of all released Fe(II) while the Fe(II)5minFZ and Fe(II)aq made up ∼10% each. In the fine fraction all components of labile Fe(II) were equal in magnitude within the experimental uncertainty. This result may be a function of the surface area of the particles, the mineralogical composition of the phases present, and/or the presence of species which may reduce Fe(III) to Fe(II) over the course of the 22-hour extraction experiment.
 Coarse- and fine-Fe(II)total,5minFZ are plotted in Figure 4c in stacked bars, together with the total Fe concentration. No statistical difference could be detected between the group 1 and group 2 Fe(II) data. Figure 4c shows that most (87.2 ± 6.1%) of the Fe(II)total,5minFZ, which averages 9.76 ± 3.37 ng m−3, is present in the fine fraction. Compared to the total Fe as determined by ICP-MS, the combined coarse and fine Fe(II)total,5minFZ account for only 1.3 ± 0.5% (geometric mean), ranging from 0.7 to 2.9%. The average coarse-Fe(II)total,5minFZ concentration of 1.16 ± 0.55 ng m−3 amounts to 0.3 ± 0.1% (geometric mean) of the coarse Fe, whereby the fine-Fe(II)total,5minFZ concentration of 8.60 ± 3.29 ng m−3, amounts to 2.4 ± 1.2% (geometric mean) of the fine Fe. As mentioned earlier, note that total Fe is almost evenly distributed between the fine (52.8%) and the coarse (47.2%) fractions.
 Compared to previous observations reported by our group, the present samples show relative Fe(II) abundances that are slightly larger. During the inter-monsoon of 1995 [Siefert et al., 1999] an average of 0.3% of the total Fe was observed as Fe(II), whereby the actual combined coarse and fine Fe(II)total,5minFZ amounted to 5.2 ± 4.4 ng m−3. During the Atlantic Ocean cruise [Johansen et al., 2000], the measured Fe(II)total,5minFZ concentrations averaged 3.14 ± 1.35 ng m−3, which accounted for 0.5 ± 0.4% of the total Fe. Thus the Fe(II)total,5minFZ concentrations reported here are larger by a factor of 1.9 and 3.1, respectively, compared to those we previously reported. This observation may be indicative of the anthropogenic constituent playing an indirect but pronounced role in the speciation of iron.
 Few studies exist on Fe(II) concentrations measured over remote oceanic regions. Zhu et al. [1993, 1997] reported similar concentrations to those found in the present study, while Zhuang et al.  measured relative Fe(II) values that were considerably larger (15%, corrected value [see Zhu et al., 1993]) in Barbados. However, these large discrepancies may be due to different sample handling and more vigorous experimental extraction procedures.