Air was isokinetically sampled about 10 m above sea level in front of the ship's stack. Only air masses without contamination, indicated by number of particles and relative wind speed and direction, were sampled. To this end, sampling was automatically switched off when particle number concentration rapidly increased, the relative wind direction was forward of the beam or the relative wind speed was below 3 m/s. The air was heated slightly (2–10°C) to sample air at a constant level of 55% RH. For a detailed description of the aerosol sample inlet see Quinn et al. [2002a].
 Three-stage multijet cascade impactors [Berner et al., 1979] were used to determine carbon and organic species. The 50% aerodynamic cutoff diameters, Dp, were 10, 1.1, and 0.18 μm, respectively. Thus, supermicrometer particles were collected in the size range 1.1 μm < Dp< 10 μm and submicrometer particles in the range 0.18 < Dp < 1.1 μm. The sampling flow rate was 30 l/min.
 Aluminum foil was used to sample for the determination of organic and elemental carbon (OC/EC). Tedlar film was used as the substrate for the subsequent analysis of organic ions by capillary electrophoresis (CE). In the latter case Pyrofoil® covered part of the Tedlar film in order to determine semivolatile organic species by Curie point pyrolysis-gas chromatography/mass spectrometry (CPP-GC/MS) from the same sample.
 The impactors sampling for the analysis of carbon were equipped with two quartz fiber filters in series behind the impaction plates to account for particles with Dp < 0.18 μm. The second filter should account for positive artifacts due to adsorption of gas phase organics. However, during this campaign the difference of the two filters was in the range of field blanks. Consequently, values of the backup filters are disregarded. The error caused by this procedure is small as can be concluded from the carbon size distributions discussed below.
 In total, 41 samples for the determination of carbon have been taken continuously in time intervals of 12–48 h, depending on the available aerosol concentration. Mean concentrations are reported as time-averaged values. 7 samples have been analyzed for organic species.
 Furthermore, a 7-stage multijet cascade impactor (50% cutoffs: 0.18/0.3/0.54/1.1/2.0/4.1/10.3 μm) was used to obtain three examples of the size distribution of carbon.
 Carbon was determined in this study separating OC/EC with a thermographic method using a commercial carbon analyzer (C-mat 5500, Ströhlein, Germany) consisting of a free programmable combustion furnace (IR 05) followed by a resistance oven (D03 GTE) holding the CuO catalyst (to convert carbon quantitatively to carbon dioxide) at 850°C, and a NDIR detector measuring the IR absorption of the formed carbon dioxide. Analyses were carried out placing samples in a quartz tube positioned in the center of the IR oven and heating rapidly under different conditions. As carrier gases in a first step nitrogen (5.0) for OC volatilization at 590°C and in a second step oxygen (4.5) for EC combustion at 650°C were applied. In between the two runs the IR furnace was cooled down to 50°C to avoid EC losses during flushing with oxygen. The duration of each run was 8 min and thermograms of every run were stored in a computer. Quantification was performed by integration of the area below the thermogram curve.
 The OC-defining temperature of 590°C was chosen with respect to the same Curie point temperature of the sampling foils employed simultaneously to detect organic single species by CPP-GC/MS (see beneath). Aluminum foils are the most appropriate substrate for impactor sampling (carbon-free production process resulting in low blank values, cheap). The melting point of 659°C restricts the EC determination to this temperature. However, tests with carbon black and natural soot (sampled on quartz filters) show that only 2–3% of the TC evolves between 650°C and 850°C. Thus, EC (and TC) does not contain carbonate carbon evolving at temperatures > 650°C [Petzold and Nießsner, 1996].
 Calibration of the instrument was performed using external standards (e.g., a solution of potassium hydrogen phthalate containing 10 μg C in 100 μl solution). Fivefold EC analysis of 62 high volume samples result in a coefficient of variation of 4.7%.
 The detection limit of the carbon analyzer was determined to be 0.3 μg carbon. Several field blanks have been taken by handling the aluminum foils in the same matter as the sampled substrates, but without drawing air through the impactors. Field blank mean values (averaged over 9 blanks) were 5.8 ± 1.1 μg and 1.3 ± 0.5 μg for OC and EC, respectively, resulting in detection limits (95% confidence level) of 0.16 μg/m3 and 0.07 μg/m3 for OC and EC, respectively, for a typical day/night sampling period. Since the sampling period was extended up to a whole day (in one case up to two days) for clean air masses the detection limit is proportionately lower. These values can be taken as uncertainty for the chemical analysis, apart from below mentioned OC/EC-split. Sampling artifacts are unknown and therefore have not been corrected for.
 OC/EC-separation can be carried out in many different ways. Methods have been developed including extraction steps to remove OC with organic solvents, acids, bases and water [Cadle and Groblicki, 1980; Japar et al., 1984; Kuhlbusch, 1994; VDI 2465 Bl.1, 1996]. Dry thermal desorption methods under an inert gas [Malissa et al., 1976; Cadle et al., 1980; Ogren et al., 1983; Ulrich et al., 1990; VDI 2465 Bl.2, 1997], air [Cadle and Groblicki, 1980; Cadle et al., 1981] or oxygen atmosphere [Malissa et al., 1976; Novakov, 1982; Cachier et al., 1989; Lavanchy et al., 1999] with a wide range of conditions also have been applied. Furthermore, combinations of both wet and dry methods have been used [Iwatsuki et al., 1998; Zappoli et al., 1999]. In addition, optical pyrolysis correction procedures have been developed and incorporated in existing instruments [Johnson et al., 1981; Huntzicker et al., 1982; Japar et al., 1984; Chow et al., 1993].
 With respect to the highly desirable comparability of carbon measurement values from different research groups inter-laboratory method comparison studies were performed [e.g., Cadle and Mulawa, 1990; Countess, 1990; Shah and Rau, 1991]. Our laboratory participated in the Germany-internal VDI/DIN-comparison experiment for EC determination [Neuroth et al., 1999] and in the International Aerosol Carbon Round Robin Test Carbon Shoot Out Stage I [Schmid et al., 2001] as well as Stage II [manuscript in preparation].
 All experiments have shown sufficient comparability of TC values but a wide variety in the results of OC and EC determinations. A survey of the results seems to show a tendency for higher EC values by pure thermographic methods resulting in lower OC. However, the basis in these comparisons is the mean value of all participants, the true value remains unknown.
 Compared to extraction methods thermal desorption methods are easier to practice, in fact they seem to be the only reliable method for impactor samples since extraction of impactor foils bears the risk of washing away part of the sample.
 With thermal desorption methods a main difficulty consists of the definition of an OC/EC separation point that enables complete OC volatilization without EC losses or EC artifact formation by charring processes during volatilization.
 From test series with organic compounds relevant in atmospheric processes it was evident that 590°C over 8 min may not be sufficient for quantitative volatilization in all cases. For methanesulfonate (MSA) or dicarboxylic acids losses of 5–20% were observed, whereas PAHs like pyrene or benzo[ghi]perylene were found to be volatilized quantitatively. Higher temperatures up to 850°C led to higher OC yields indicating that the incomplete OC determination was not only caused by charring processes during heating but also by incomplete volatilization. Detailed results will not be presented here since the degree of volatilization of ionic compounds (accounting for the least volatile species) strongly depends on total ionic composition. For example, oxalic acid, a major aerosol constituent (see below), will dissociate in wet particles and, in the presence of calcium (as part of sea salt), possibly form the less soluble calcium oxalate. This species decomposes by heating to form CaCO3 (which is stable against further thermal decomposition up to 650°C and higher) and CO (which will be measured as CO2). Thus, in the case of total transformation to calcium oxalate only 50% of oxalic acid would be determined as OC. Comparable reactions might occur for other ionic compounds. Such phenomena have been observed in the analysis of atmospheric aerosol particles [Novakov and Corrigan, 1995].
 With respect to oxidizing methods for the separation of OC and EC the volatilization procedure used here has the important advantage of being short in time (8 min versus 1–3 h) which makes it suitable for the large number of samples resulting from size-fractionating and long term analysis. Nevertheless, charring is expected to be higher. This means that the method applied here provides a lower limit for OC and upper limit for EC. Higher OC volatilization temperatures result in higher OC/EC ratios. For some samples in central Europe the difference between 590°C and 650°C has been determined to be in the range of 15%. However, an upper temperature limit of about 650°C due to the melting point of aluminum exists.
 In summary, there is no generally accepted procedure for OC/EC separation until now, all results are depending of the method applied and therefore results from different experiments are difficult to compare. The method used here leads typically to lower OC/EC ratios compared to related techniques. However, the lack of positive sampling artifact and the short analysis time provides are main advantages of the applied methods. The OC/EC determination development and characterization will be discussed elsewhere in more detail [Gnauk et al., in preparation].
 To convert organic carbon (OC) to organic matter (OM) a conversion factor of 1.6 is used. This value results from an estimation for strongly anthropogenic influenced aerosols based on a recent study by Turpin and Lim . Although the specification of organic material that the cited study is based on was incomplete, such conversion factors are necessary to derive mass fractions of organic material. The results discussed in this paper show that during the INDOEX campaign highly oxidized material (i. e. carboxylic acids that would result in a large conversion factor) was found to a much larger extent than low oxidized material (i. e. alkanes, PAHs etc. that would result in low conversion factors). Thus, the use of a factor of 1.6 is expected to be a conservative estimation of organic matter.
 The absorption for submicrometer and supermicrometer particles was determined at 550 nm and 55% RH by measuring the change in transmission through a filter with a Particle Soot Absorption Photometer (PSAP, Radiance Research). Measured values were corrected for scattering, spot size, flow rate and the manufacturer's calibration according to Anderson et al.  and Bond et al. .
2.4. Organic Specification
 Aliphatic and hydroxylated organic acids as well as semivolatile compounds were determined using a combined method of capillary electrophoresis (CE) and Curie point pyrolysis-gas chromatography/mass spectrometry (CPP-GC/MS). Both methods require only small sample amounts and avoid extensive purification or derivatization procedures.
 Briefly, for determination of organic acids the Tedlar films were cut into small pieces and leached into 0.75 mL pure water. After filtration the solutions were analyzed by capillary zone electrophoresis with indirect UV detection. A p-amino-benzoic acid based electrolyte (pH 9.5) and detection at 254 nm is applied. Quantification is performed by comparing the peak areas with those of external standards.
 Semivolatile compounds were determined by heating the exposed Pyrofoil® (Fe-Ni-alloy) rapidly due to absorption of a radio frequency magnetic field by the ferromagnetic material. At the Curie point temperature (here 590°C), which is reached within milliseconds, the ferromagnetism changes to paramagnetism and the heating effect ceases. The volatile and semivolatile organic compounds evaporate from the foil and are flushed into the GC by helium. The substances were separated by a CP-Sil-5 capillary and identified by a quadrupol mass spectrometer with an electron impact ionization source (Trio 1000). The quantification is performed by comparing the peak areas with those of internal deuterated standards and calculating response factors with external standards, respectively. The methods are described in detail in the work of Neusüß et al. [2000a].
 Particle mass was obtained by weighing the substrates before and after sampling. A Cahn Model 29 microbalance was used at a relative humidity of 33 ± 3%. Thus, the mass includes water associated with the sampled particles at 33% RH. A detailed description of the weighing procedure is given by Quinn and Coffmann .
 The weighing was performed for samples taken by a two stage impactor resulting in mass concentrations for supermicrometer particles (Dp = 1.1–10 μm) and submicrometer particles (Dp < 1.1 μm). Therefore, the submicrometer mass includes particles with Dp < 0.18 μm, in contrast to the submicrometer carbon mass. From mass size distributions with 7-stage impactors (as for the size segregated carbon with a filter behind the last stage) a factor converting the mass of particles with Dp < 1.1 μm to the mass of particles with Dp = 0.18–1.1 μm can be derived. This factor was determined to be 0.95 ± 0.01 and 0.85 ± 0.05 for polluted and clean air masses, respectively, and was used for correction.