Carbonaceous aerosols over the Indian Ocean during the Indian Ocean Experiment (INDOEX): Chemical characterization, optical properties, and probable sources

2.1. Sampling

[7] Aerosol sampling was conducted on the NCAR C-130 aircraft, based on Hulule Island (4.2°N, 73.5°E), Republic of the Maldives, during INDOEX IFP in February−March 1999 cf. C-130 overview paper by [Clarke et al., 2002]. A total of 18 research flights (RFs) were performed over the Indian Ocean region covering a range in altitudes from 0 to 6.5 km. Here, we present results from 13 of these flights.

[8] Samples were collected through the community aerosol inlet (CAI) of the aircraft. The CAI, about 6.7-m total length, is mounted on the side of the aircraft and projects ahead of the aircraft nose. It supplied sample air to multiple instruments, which subsampled isokinetically via individual sampling tubes of different diameters. The 50% cut-size of the CAI is about 3 μm; and for particles near 1 μm, its efficiency is about 90% in the moist boundary layer (better for dry aerosol aloft) (A. Clarke, personal communication, 2000). A detailed description is given by Blomquist et al. [2001]. Temporal resolution for the collection of our samples was about 10 to 30 min in the boundary and pollution layers, and up to two and three hours in the free troposphere. The relative humidity (RH) averaged 70 ± 17% in the MBL and 30 ± 16% in the rCBL legs.

[9] Aerosol samples were collected on two stacked-filter units (SFUs) mounted in parallel (Figure 1). The SFUs were connected to one CAI pick-off tube of 1.8-cm i.d. They each consisted of two polyethylene 47-mm diameter filter holders (NILU) connected in series. During operation, a Nuclepore prefilter (PC Membrane, Corning Costar, nominal pore size 8.0 μm) was placed in the first stage of each unit. This filter collected the coarse fraction which consisted of particles larger than ca. 1.3 μm (calculated for an average face velocity of 47 cm sec⁻¹ in the boundary layer [John et al., 1983]), and smaller than 3 μm (50% size-cut of the CAI) aerodynamic diameter.

[10] One SFU was used for the analysis of the carbonaceous component of the aerosols. In this unit, two quartz filters (Pallflex Tissuquartz 2500 QAT-UP) were placed directly on top of each other (tandem arrangement) in the second stage. The pressure drop across the tandem filter arrangement, for our face velocity of 47 cm sec⁻¹, was 25 inches of H₂O. This was calculated based upon the published change in pressure of 0.0013 atm per cm sec⁻¹ reported by McDow and Huntzicker [1990] for a tandem quartz filter arrangement. Before use, the quartz filters were baked at 800°C for 12 h to remove residual organic impurities. As discussed below (9), the tandem arrangement was used to correct for the adsorption of gaseous organic compounds by the filter material, commonly referred to as the positive artifact [Kirchstetter et al., 2001; Turpin et al., 1994]. If not accounted for, this artifact results in an overestimation of organic aerosol mass concentrations.

[11] The second SFU was used for the analysis of water-soluble ions. The second stage of this unit had a PTFE Teflon membrane filter (Zefluor Membrane filter, Pall Gelman; nominal pore size 2.0 μm, (D_p < ∼1.2 μm)) to collect the fine aerosol fraction. Details of the analysis of this aerosol component are discussed by Gabriel et al. [2002].

[12] Loading and unloading of the filter units was performed before and after every flight in the laboratory facilities at Hulule Island. After the unloading process, filter samples were stored in a freezer at −18°C in precleaned glass vials until analysis. Vials were precleaned according to the procedures recommended by Salmon et al. [1998].

[13] Blank filters were collected on every flight using filter units loaded in the Hulule laboratory. These units were connected to the aerosol inlet in the same manner as the real samples, with flow being applied for only 15 s.

2.2. Optical Measurements

[14] A three-wavelength nephelometer (Model 3563, TSI Incorporated, St. Paul, Minnesota USA), and a continuous light absorbing photometer (particle-soot absorption photometer, PSAP, Radiance Research Inc., Seattle, WA, USA) were operated at a RH of about 40% and used an impactor that allowed collection of particles of aerodynamic diameter less than 1 μm or 10 μm. The nephelometer measured dry aerosol total scattering (σ_sp) and hemispheric backscattering of light at three visible wavelengths: 450, 550 and 700 nm. Here, we report σ_sp for particles of D_p < 1 μm at 550 nm. At this wavelength, the detection limit for 30-second averages is 1 Mm⁻¹. The PSAP measured particle absorption (σ_ap) continuously via attenuation of 565-nm radiation by particles collected on a quartz filter. The PSAP data were adjusted to a wavelength of 550 nm using the correction scheme described by Bond et al. [1999]. PSAP results presented here also correspond to particles of D_p < 1 μm. The detection limit of this instrument is less than 1 Mm⁻¹ for 1-min averages. Detection limits of both instruments decreased with longer averaging times; therefore, in our particular case (legs ≥ 10 min), these are conservative estimates. Sheridan et al. [2002] explain the operation of these instruments and discuss the uncertainties of σ_sp and σ_ap measurements.

2.3. Analysis

3.1. Interpretation of Thermograms

[18] Most of the INDOEX C-130 samples were collected from heavily polluted air masses. Their thermograms showed similar profiles, with carbon evolution extending to 600°C (Figure 2). Samples collected south of the Intertropical Convergence Zone (ITCZ) resulted in relatively featureless thermograms (not shown here), most likely due to the pristine conditions existing there.

[19] In general, thermograms of front filter samples collected in polluted air masses consist of several discrete peaks. For example, those of the front filters shown in Figure 2 consist of four peaks (i.e., A, B, C, and D). The first peak (A), representing the most volatile material in the sample, has a maximum intensity around 100°C and extends to ∼200°C. Frequently, thermograms of backup filters (solid squares in Figure 2) show a peak at the same temperature and of similar intensity. The front filter removes all particles from the sampled air stream. Therefore, the origin of peak A on the backup filter is the adsorption of organic compounds that were originally present in the gas phase, and which adsorb to the front and backup filters during sampling [Kirchstetter et al., 2000]. Sometimes, inhomogeneity in the filter stock could produce differences between the intensities of the volatile peak in front and back filters [Kirchstetter et al., 2001]. In such cases, peak A is not included in the integration of the thermogram. The second peak (B) in Figure 2 is completely absent from the back filter, indicating that it is in the particulate phase. Similarly, peak C is more prominent in thermograms of front filters than back filters, indicating the collection of particulate-phase OC. Peak C in thermograms of backup filters may have been the result of adsorbed gases or low-level contamination.

[20] In environments influenced by combustion sources, such as the INDOEX region, the particle-phase carbonaceous material consists of both OC and BC. The high-temperature thermogram peak (D), which was centered at about 500°C and which is not present in thermograms of backup filter samples in the majority of cases, is used to estimate the amount of BC in the sample. To calculate the area under peak D (BC peak), we integrated the carbon evolving at temperatures above the minimum between peaks C and D, when these two peaks where well separated. In cases where these two peaks were not so well resolved, we looked at the shape of peak D and if it showed to be symmetric (bell-shaped), then we integrated the higher temperature half of the peak and mutiplied this area by 2 to get the total area. For cases where peaks C and D were not resolved enough to apply one of these 2 approaches (about 7% of the total number of samples), BC was not determined. It has been shown that the carbonaceous material giving rise to this high-temperature peak is largely responsible for visible light absorption [Gundel et al., 1984].

[21] Based on the information obtained from the thermograms, it is possible to estimate the concentrations of the carbonaceous components (i.e., TC, BC, and OC) in the samples. The mass of carbonaceous material on the backup filter is subtracted from the mass of carbonaceous material on the front filter to correct for the adsorption artifact (assuming that the contribution of gas-phase organics, peak A, is the same on both filters). This difference is defined as total carbon aerosol, TC, and reflects the mass of particle-phase carbonaceous material collected on the front filter. The OC is estimated as the difference between TC and BC.

[22] One uncertainty in the determination of BC and OC concentrations is the possibility that OC may char (undergo pyrolysis) during thermal analysis, and then coevolve with BC. Similarly, highly refractory organic material in the sample may be indistinguishable from BC. Both occurrences would lead to an overestimate of BC concentration. These factors led Gundel et al. [1984] to conclude that a combination of thermal, optical, and solvent extraction methods must be used to accurately estimate BC. Optical (light-absorption) and solvent extraction (treatment to remove organic material from the sample before it is analyzed for BC) methods were employed for only a small number of INDOEX samples. These few results indicate that BC determination was not affected by coevolving organic material. Furthermore, a positive characteristic of the EGA method chosen for this study is the use of pure oxygen as the carrier gas. Compared to similar thermal techniques that expose the sample to inert gas (nitrogen or helium), the use of pure oxygen minimizes charring of organic material [Lavanchy et al., 1999].

[23] Many of the samples analyzed in this study produced a thermogram peak that evolved beyond the temperature normally associated with BC. Carbonate carbon, which is usually not encountered in submicron samples, has an evolution temperature greater than that of BC. Powdered coral (calcium carbonate), for example, produces a single thermogram peak centered at 655°C. While the literature states that exposure to hydrochloric acid vapor will rid the sample of carbonate [Chow et al., 1993], this treatment did not remove, or even reduce, the magnitude of the high-temperature peak in question. As a result, it is not certain what material gives rise to this peak. We believe, however, that because this peak was found in thermograms of backup filters and blanks as well as front filters it is indicative of contamination and was not collected during sampling. Thus, the thermogram area beyond the evolution of BC was not included in the estimate of carbonaceous aerosol concentration.

3.2. Estimation of Particulate Organic Matter

[24] The OC determined with thermal methods such as EGA is the carbon mass concentration in the particulate organic matter (POM). A POM/OC conversion factor is frequently used to estimate POM mass from the OC content. However, this conversion is at best approximate because the molecular-to-carbon mass ratio is not precisely known [Turpin and Lim, 2001]. As reported by Turpin and Lim, these values can range from 1.2 to as high as 2.6, depending on the aerosol composition. This uncertainty could lead to errors in the determination of POM and, therefore, the aerosol mass. The factor used for our INDOEX samples was 1.7, which corresponds to an empirical formula of ca. CHO_0.5, since the sampled air masses were strongly influenced by anthropogenic emissions from fossil fuel and biofuel combustion. This POM/OC factor assumes that the organic aerosol species, in cases like this, will be less oxidized than in air masses influenced by natural organic aerosols.

3.3. Pollution Layers and Air Mass Back Trajectories

[25] During most of the INDOEX IFP C-130 flights, two polluted layers were observed within the boundary layer (Figure 3). Vertical profiles of σ_sp and dew point temperature along with the concentrations of sodium in the coarse fraction (which can be a tracer for sea salt in the MBL) were used to define these layers. The first one was the well-mixed marine boundary layer (MBL) reaching up to ∼1.2 km. The second one was a residual continental boundary layer, rCBL, between 1.3 and ∼3.2 km. This rCBL was formed when continental air masses originating from the Indian subcontinent were advected over the relatively cold ocean. The weaker convective activity over water could not sustain vertical mixing to the same height as over land. Therefore, a new MBL evolved at the base of the CBL, and the rCBL (between the top of the new MBL and the inversion at the top of the CBL) became disconnected from the free troposphere and the sea surface. More details about the mechanism of formation of the rCBL are presented by Johnson et al. [2000].

[26] Rasch et al. [2001] and Verver et al. [2002] present analyses of the meteorological conditions during the INDOEX IFP period. These analyses indicate that there was a significant change in the circulation patterns of the air masses between February and March. They divided the total period in two halves. The first half is what we call our “February period” (in which we include RFs 2 to 10). Here, air masses at surface levels in the Indian Ocean region were mainly coming in the northeast (NE) trades over the Western Bay of Bengal and/or in a flow from the NE, out of Southeast Asia (SEA). The second half is the “March period” (here, we include RFs 12, 13, 15, and 18) where air was coming mainly from a flow off the West Coast of India (the plume from Asia, SEA channel, was absent from March 1–22). These analyses were mostly limited to the near-surface level and/or below 2.5 km. This limits our knowledge in terms of the meteorological conditions prevailing for the rCBL. Also, RF 18 was not included in the analysis made by Rasch et al. [2001], since it was made on 25 March and their analyses ended on 22 March.

[27] Three major source regions were derived from the BT analyses: (1) India (INDIA), (2) Southeast Asia (SEA), and (3) West Asia (WA) (Table 1). Some samples were also collected in air masses that did not make contact with land over the past 10 days. They are classified as coming from the North Indian Ocean and/or the Arabian Sea (NIO), or coming from the South Indian Ocean (SIO). Flow from the northeast was evident in most of the trajectories, with most of the air masses originating from the Indian and south Asian subcontinents [Rasch et al., 2001; Verver et al., 2002]. The dominant source of aerosols in the Arabian Sea, the Bay of Bengal, and the NIO near the surface was India. At higher altitudes (above 1.2 km), contribution from other countries such as Asia, Arabia, and Africa was more evident, as suggested by the BTs and the analyses by Rasch et al. [2001].

[28] Results reported in Table 1 are classified according to the tropospheric layers sampled (MBL or rCBL), the period that they represent (February or March), and the source region of the sampled air mass (INDIA, SEA, WA, NIO, SIO). Results from the samples collected in the free troposphere (>3.2 km) are not presented here. Table 2 provides a summary of the averaged values for the different classifications (i.e., layers, periods, and sources) and contains values from other studies for the purpose of comparison.

3.4. Mass Concentrations

[29] Significant variability between the mass concentrations of TC aerosol (from now on TC), BC, and OC was observed between samples from the same layers, periods, and/or source region, as it is shown in Table 1. However, the regionally averaged concentrations (INDIA, SEA, WA) for these species (Table 2 and Figures 4a and 4b) show low variability between them. Rasch et al. [2001] explain that the long residence times of aerosols (about 7 to 8 days for carbonaceous aerosols and sulfate) and the low wet deposition removal rate due to lack of rain during the winter/spring monsoon may be responsible for the great extent of polluted air masses over a large domain of the INDOEX region. These factors, assuming well-mixed layers, also could have resulted in the low regional variability in the average concentrations for the species under study. The following discussion pertains mainly to the samples collected in polluted air masses, below 3.2 km (in the MBL and/or rCBL), excluding the samples coming from oceanic regions (i.e., SIO and NIO) since these air masses were, in general, comparably unpolluted.

[30] Total aerosol mass in this study was calculated by summing the individual species measured (i.e., BC, POM (OC' × 1.7), and the water-soluble ions) and does not include insoluble inorganic components (fly ash and mineral dust). OC' represents the OC after the subtraction of the carbon mass of the organic species determined by ion chromatography (i.e., oxalate, formate, and MSA). We refer to the total aerosol mass as “estimated aerosol mass” (EAM). The EAM ranged from 3.2 to 41.3 μg m⁻³, with averages of 24.2 μg m⁻³ and 13.8 μg m⁻³ for the rCBL and the MBL, respectively. The average for the total period (MBL and rCBL) was 15.3 μg m⁻³. These values are comparable to pollution levels observed in industrialized regions like the mid-Atlantic coast of the United States (TARFOX experiment), where the aerosol mass observed was on average about 15 μg m⁻³ [Hegg et al., 1997], and in the Negev Desert in Israel (ARACHNE experiment) (highly influenced by anthropogenic sources) where this value was about 15.9 μg m⁻³ [Andreae et al., 2002]. Figure 5 shows the contributions of aerosol species to EAM during the total INDOEX IFP. Major fractions were POM (35%), SO₄²⁻ (34%), BC (14%), and NH₄⁺ (11%). The contribution of the other water-soluble ions analyzed was about 6%.

[31] The uncertainty in the value of the POM/OC factor (see 10) could have led us to an underestimation of the POM and, therefore, of the EAM for some of our samples. The magnitude of the error in these results is difficult to estimate due to the lack of knowledge about the appropriate POM/OC factors. However, if instead of 1.7, we apply 2.6, one of the highest values reported for POM/OC (for an aerosol heavily impacted by wood smoke) (Turpin and Lim, submitted manuscript, 2000), to the mass concentrations of OC, the POM values would increase by a factor of 1.5. As a result, the EAMs would increase by about 10–16%, depending on their OC content. Since the INDOEX aerosol also has a contribution from fossil fuels, however, the POM/OC factor is unlikely to be greater than two, and consequently, EAM errors due to uncertainties about POM/OC are probably less than 10%.

[32] Concentrations of TC, BC, OC, and POM for the different periods and layers are presented in Tables 1 and 2. TC represents about 37% of the EAM. Table 2 shows that TC concentrations were much higher than in continental air masses during ACE-2 [Novakov et al., 2000a], but comparable to the values obtained during TARFOX [Hegg et al., 1997], suggesting very high levels of pollution. This was also supported by the high concentrations of BC (up to 6.3 μg C m⁻³) and OC (up to 9.4 μg C m⁻³) observed.

[33] The t-test was applied to the concentrations of TC, OC, BC, and EAM, and the differences between the February and March periods were found to be statistically significant only for BC (p = 0.03). It was found, however, that there were significant differences between the layers, especially during the February period, with the tendency for higher EAM and carbon fractions in the rCBL during the February period (Figure 6). The lowest EAM was found for particles measured in air masses coming from the WA region (Figure 4c). This is also supported by the results presented by Cautenet et al. (unpublished manuscript, 2002) whose model predicted the lowest concentrations of BC in the western part of the Indian Ocean. Removal processes, mainly dry deposition, as well as low amount of biofuel in comparison to SEA and INDIA could have been the reason for these lower aerosol burdens. For the SEA and INDIA regions, as well as for the layers within the March period, concentrations show low variability.

[34] It is also interesting to note that in the rCBL the %POM value (therefore, also %TC) was higher than the %SO₄²⁻ value (Table 2). Similar results were found for INDOEX-C-130 filter samples analyzed by scanning electron microscopy (M. Posfai, personal communication, 2000). Here, it was found that sulfate/soot particles were dominant in samples from RF 10 at 30 m (MBL), while at 1.5 km (rCBL), the majority of the particles were carbon spherules. During TARFOX (altitudes 0 to about 4.8 km), Novakov et al. [1997b] observed an increase in the carbonaceous fraction with increasing altitude. At even higher altitudes (upper troposphere) Murphy et al. [1998] also observed that aerosols contained more organic material than sulfate. In our case, this effect could have been due to a faster conversion of SO₂ into SO₄²⁻ in the MBL. This is supported by the results obtained on the R/V Ronald Brown (M. Norman, personal communication, 2000) where it was found that SO₂ represented only about 4% of the total sulfur (SO₄²⁻ plus SO₂) in the MBL, suggesting that most of the SO₂ had been converted into SO₄²⁻. In contrast, a substantial fraction (up to about 65%) of the total sulfur was still present as SO₂ in the rCBL [Reiner et al., 2001]. This finding highlights the importance of airborne studies, since ground-based measurements could significantly underestimate the contribution of carbon species to the column aerosol mass budget.

[35] Compared to the regions north of the ITCZ, the region south of the ITCZ was relatively unpolluted. Concentrations of TC, BC, and OC south of the ITCZ were often near or below the limit of detection (LOD, 0.2 to 0.8 μg C m⁻³) and, therefore, only a few results were obtained for this region (SIO in Table 1). The decrease in the carbon concentrations when approaching the Southern Hemisphere can be observed from the latitude profile of TC and BC over the Indian Ocean (Figure 7). High OC mass concentrations were concentrated in the Northern Hemisphere and were of the same magnitude as those of BC, in many cases, suggesting a primary origin for these aerosols. The influence of combustion-derived pollution is seen down to 5°S (north of the ITCZ). However, south of the ITCZ, where the BC reaches LOD values, the OC concentrations were low (i.e.,<3 μg C m⁻³). Therefore, we can infer that most of the OC north of the ITCZ has an anthropogenic origin.

3.5. Optical Properties of Carbonaceous Aerosols

[36] In this section, we compare our filter-based measurements with the scattering and absorption coefficients that were measured aboard the C-130 aircraft. To facilitate this comparison, the optical coefficients were averaged over the periods during which the filter samples were collected. A complete discussion of the entire optical-property data set is presented by Sheridan et al. [2002]. We note here that the two data sets (chemical versus optical results) were obtained using different particle size cuts (1.3 versus 1.0 μm) and sample relative humidities. However, preliminary analyses suggest that these differences introduce negligible error (<1%) into the results discussed below (S. Howell, personal communication, 2000).

3.5.1. Specific Absorption Cross Section

[37] Black carbon is the main light-absorbing component of aerosol particles [Rosen et al., 1978; Gundel et al., 1984; Lindberg et al., 1993]. An important optical parameter is the specific absorption cross section, α_ap (m² g⁻¹). Here we relate BC mass concentrations (in μg C m⁻³) to the measured absorption coefficients, σ_ap (m⁻¹), to derive α_ap values for INDOEX aerosols. The data (Tables 1 and 3) show that σ_ap values for this subset of INDOEX samples ranged from 1.2 to 38 Mm⁻¹, with an average for the polluted layers of 16 Mm⁻¹. These values are considerably higher than those measured in other studies over polluted regions such as TARFOX and ARACHNE (i.e., 3 ± 3 and 9 ± 4 Mm⁻¹, respectively). The high INDOEX values demonstrate the highly light-absorbing nature of this aerosol. Figure 8 shows two distinctly different relationships of σ_ap versus BC concentrations below and above about 4 μg C m⁻³. As can be seen from this figure, the scatter in the data increased significantly for points with BC > 4 μg C m⁻³. At present, we do not know the exact reason for these differences.

Table 3. Averages of Optical Aerosol Parameters (σ_ap, σ_sp, ω_o, α_ap, and α_sp) During INDOEX Filter Measurements

Source	σapa, Mm−1	σsp, Mm−1 λ = 550 nm	ωo	αap, m2 g−1		αsp, m2 g−1 @ 550 nm	References
				(BC<4.0 μg m−3)	(BC > 4.0 μg m−3)
INDOEX total	16 ± 8	62 ± 34	0.81 ± 0.05	8.1± 0.7 (r2 = 0.7)	8.3 ± 2.4 (r2 = 0.4)	4.9 ± 0.4 (r2 = 0.8)	this study
INDOEX rCBL	20 ± 9	85 ± 50	0.80 ± 0.03	7.2 ± 1.8 (r2 = 0.6)	nd	4.6 ± 0.7 (r2 = 0.9)	this study
INDOEX MBL	15 ± 8	56 ± 26	0.81 ± 0.05	7.4 ± 0.8 (r2= 0.6)	nd	4.5 ± 0.5 (r2 = 0.7)	this study
INDOEX Feb	18 ± 8	68 ± 33	0.80 ± 0.04	10.8 ± 1.6 (r2 = 0.5)	7.7 ± 2.8 (r2 = 0.1)	4.6 ± 0.7 (r2 = 0.5)	this study
INDOEX March	12 ± 8	55 ± 36	0.82 ± 0.05	5.6 ± 0.8 (r2 = 0.7)	nd	4.8 ± 0.4 (r2 = 0.9)	this study
INDOEX Feb rCBL	20 ± 10	80 ± 49	0.79 ± 0.03	nd	nd	nd	this study
INDOEX Feb MBL	18 ± 7	65 ± 29	0.80 ± 0.05	10.8 ± 1.6 (r2 = 0.5)	7.7 ± 2.8 (r2 = 0.1)	3.7 ± 0.6 (r2 = 0.5)	this study
INDOEX March rCBL	20 ± 10	91 ± 55	0.81 ± 0.01	7.2 ± 1.9 (r2 = 0.6)	nd	4.5 ± 0.0 (r2 = 1)	this study
INDOEX March MBL	9 ± 5	42 ± 12	0.83 ± 0.06	5.6 ± 1.2 (r2 = 0.6)	nd	3.2 ± 0.9 (r2 = 0.3)	this study
INDOEX INDIA	18 ± 8	72 ± 34	0.80 ± 0.04	8.0 ± 1.1 (r2 = 0.5)	7.2 ± 0.9 (r2 = 0.6)	4.9 ± 0.5 (r2 = 0.8)	this study
INDOEX SEA	15 ± 6	53 ± 22	0.78 ± 0.02	nd	nd	nd	this study
INDOEX WA	13 ± 7	55 ± 27	0.82 ± 0.05	8.4 ± 1.5 (r2 = 0.8)	nd	nd	this study
INDOEX NOAA/CMDL (0–1 km)	14 ± 7	58 ± 26	0.81 ± 0.05	–		–	Sheridan et al., 2002
INDOEX NOAA/CMDL (1–3 km)	16 ± 10	74 ± 38	0.83 ± 0.06	–		–	Sheridan et al., 2002
ARACHNE-96	9 ± 4	90 ± 30	0.92 ± 0.03	8.9 ± 1.3		3.0 ± 1.9	Formenti et al., 2001
ARACHNE (95–97)	–	87 ± 54	–	–		5.2 ± 0.2	Andreae et al. [2002]
TARFOX	3 ± 3	44 ± 50	0.90 ± 0.09	–		2.8 ± 0.3	Hegg et al., 1997
ACE–1	–	4.1 ± 2.8	0.99 ± 0.01	–			Quinn et al., 2000
ACE–2, continental air	–	40 ± 17	0.95 ± 0.03	–			Quinn et al., 2000
HUNGARYb	8.9	93	–	–		8.3 (for sulfate)	Mészáros et al., 1998

• a Values of σ_ap for TARFOX and Hungary were at 565 nm, INDOEX was 550 nm.
• b D_p < 1 μm, rural location, winter time.
• Literature values are shown for comparison. α_ap and α_sp for INDOEX were calculated using the RMS regression analysis. 1 Mm⁻¹ = 10⁻⁶ m⁻¹. NOAA/CMDL = National Oceanic and Atmospheric Administration, Climate Monitoring and Diagnostics. Values from NOAA/CMDL are for polluted air masses north of 5°N (STP). These averages correspond to the entire INDOEX optical-property data set. nd = could not determine because no data available or not enough data points for regression analysis. – = data not available or not measured.

[38] The regression analysis of the first group of samples in Figure 8 (BC < 4.0 μg C m⁻³), indicates a good correlation (r² = 0.7) and an α_ap value of 8.1 ± 0.3 m² g⁻¹ when the y-intercept is set to zero. This α_ap value is consistent with the typical range of α_ap values (8–10 m² g⁻¹) for BC originating from high-temperature combustion sources, and is in agreement with theoretical estimates [Horvath, 1993; Liousse et al., 1993; Malm et al., 1996]. An α_ap value of 6.6 ± 0.6 m² g⁻¹ was observed when the intercept was not forced to zero. At BC concentrations close to zero, absorbing species different from BC, and possibly not accounted for by our chemical analyses, could have been present. If a root-mean square analysis (RMS) is applied to this data set, the α_ap values obtained are 8.1 m² g⁻¹and 8.3 m² g⁻¹ for BC < 4.0 μg C m⁻³ and BC > 4.0 μg C m⁻³, respectively. These are the values presented in Table 3 and they show no significant differences between the absorption properties within the two layers. However, differences between the February and March periods are noticeable, with α_ap values of 10.8 m² g⁻¹ and 5.6 m² g⁻¹, respectively. These differences could have been, for example, due to differences in the mixing state or the absorbing properties of the BC.

[39] The observed α_ap values and the good correlation between the σ_ap values and BC mass concentrations for the whole period (BC < 4 μg C m⁻³) suggest that BC is responsible for the strong light absorption observed in the dense brownish pollution haze over the Indian Ocean during INDOEX. This conclusion is further supported by the results of modeling studies [Satheesh et al., 1999; Podgorny et al., 2000].

3.5.2. Specific Scattering Cross Section

[40] The value of σ_sp is highly correlated with the mass of aerosol particles in the submicron diameter range [Waggoner et al., 1981]. Therefore, the specific scattering cross section, α_sp, can be derived from the relationship between σ_sp and the total submicron mass. Values of α_sp reported in the literature are mostly determined by considering sulfate as the main scattering species. These values range from 1.8 to 13 m² g⁻¹ [Charlson et al., 1999; Andreae et al., 2002]. However, sulfate is not the only important species when considering aerosol light scattering. Other chemical components can contribute significantly to the submicron mass of the atmospheric aerosol and, consequently, to the attenuation of visible wavelengths [White, 1990]. Fine carbonaceous aerosol is particularly important in this regard. During TARFOX, Hegg et al. [1997] demonstrated that carbonaceous aerosol can contribute, on average, about two thirds of the total dry aerosol scattering. Using their derived aerosol mass concentrations, they found a α_sp value of 2.8 ± 0.3 m² g⁻¹ (r² = 0.77).

[41] The two species that dominated the total aerosol mass in the INDOEX pollution layer were POM (35%) and sulfate (34%). In order to determine the relative contribution of these species to the σ_sp values, we attempted to use multiple linear regression analysis (MLR). However, statistical analyses showed that sulfate and POM are not statistically independent, probably sharing similar sources and sinks. As a result, it was not possible to derive separate α_sp values for these species from the MLR analysis. For this reason, we used the measured σ_sp values and the total estimated aerosol masses to determine the α_sp values (Figure 9).

[42] Dry σ_sp values for our samples ranged from 1.3 to 192 Mm⁻¹ (Table 1), with an average for the overall period of about 62 Mm⁻¹. Similar results were obtained for the entire INDOEX-C-130 data set [Sheridan et al., 2002]. The average and maximum values are comparable to σ_sp values obtained in regions influenced by anthropogenic emissions (Table 3) like the mid-Atlantic coast of the US during TARFOX (values ranged from 53–196 Mm⁻¹), the Negev Desert in Israel during ARACHNE (average was 90 Mm⁻¹), the continental air masses studied during ACE-2 (values ranged from 0.11 to 100 Mm⁻¹), and a rural site in Hungary during the winter time (average of 93 Mm⁻¹) [Hegg et al., 1997; Quinn et al., 2000; Andreae et al., 2002]. In comparison, values observed during ACE-1, where the anthropogenic influence was minimal (mainly natural maritime aerosols) were much lower (0.66 to 38 Mm⁻¹).

[43] Excluding three outliers (Table 1), for which α_sp would have physically impossible values, we found an α_sp value of 4.9 m² g⁻¹ (using RMS analysis), with a very good correlation (r² = 0.8) (Figure 9). This value is close to values estimated by Penner et al. [1998] for ammonium sulfate (5.07 m² g⁻¹) and fossil fuel OC and BC (4.84 m² g⁻¹). Also, it compares well with the value of 5.2 m² g⁻¹ obtained for the fine aerosol fraction during ARACHNE (1995–1997) [Andreae et al., 2002].

[44] The values of α_sp presented in Table 3 show no significant differences within the different layers and sampling periods, suggesting aerosols of similar size distributions and/or refractive indices.

3.5.3. Single Scattering Albedo

[45] The single scattering albedo (ω_o), which is the ratio of scattering to total extinction, was estimated from the σ_sp and σ_ap values using the equation,

All values were between 0.74 and 0.95 (for dry conditions), with an average of 0.81 ± 0.05 for the polluted layers. Similar results for the entire INDOEX data set are presented by Sheridan et al. [2002] and are included in Table 3. A comparison with ω_o values observed at other locations is also presented in Table 3. In general, INDOEX values of ω_o were significantly lower than at other locations influenced by anthropogenic emissions (Table 2) and constant within the different periods and layers. This is consistent with the unusually high fraction of BC found (14% on average), suggesting that the chemical species responsible for this large absorption is indeed BC. Other studies have reported similar results in terms of the highly absorbing aerosol present over the Indian Ocean region [Ackerman et al., 2000; Satheesh and Ramanathan, 2000].

[46] As outlined in the preceding two subsections and from Table 3, the optical properties of the INDOEX fine aerosol are not significantly different between the layers (MBL or rCBL). However, in the MBL, values of ω_o and α_ap (as mentioned before) along with the BC concentrations seem to vary between the two periods pointing to aerosols with a high absorbing character during the February period.

[47] The value of ω_o determines the sign of the radiative forcing [Haywood and Boucher, 2000]. The low ω_o values observed during INDOEX, resulting from the high fraction of BC (high absorption), result in a strongly negative radiative forcing of aerosols at the surface (−23 W m⁻²), but a positive forcing at the top of the atmosphere (+16 W m⁻²) (Ramanathan et al., unpublished manuscript, 2000). This could have significant impacts on climate. For example, a decrease in surface solar heating could decrease evaporation and thus the intensity of the hydrological cycle. Additionally, heating of the atmosphere could evaporate low-level trade cumulus clouds embedded within the aerosol layer resulting in a net warming influence [Ackerman et al., 2000; Satheesh and Ramanthan, 2000; Ramanathan et al., 2001].

3.6. Sources of Aerosols in the Polluted Layers

[48] During INDOEX IFP the dominant source region of aerosols in the study region (Arabian Sea, Bay of Bengal, and NIO) was India. In this country, for the year 1990, burning of biomass and fossil fuels contributed about 64% and 36% to the total fuel consumption, respectively. The largest contributions were from the burning of wood (33%), coal (30%), animal waste (14%), and agricultural residues (13%) [Venkataraman et al., 1999].

[49] Most of the biomass burned in India is used as fuel for cooking and heating. Streets and Waldhoff [1998] estimated this value to be about 49% of the total energy consumption, mainly fuel wood (60%), crop residues (21%), and animal waste (19%). These values are similar to the ones presented by Venkataraman et al. [1999] for biomass (wood 51%, animal waste 22%, and agricultural residues 20%). The contribution of forest fires and burning of grassland is thought to be low (less than 7%).

[50] The major fossil fuels burnt in India are coal and diesel. Venkataraman et al. [1999] estimated a contribution of 49% from coal emitted in power plants, 8% from coal from the steel industry, and 8% from diesel. For 1996, Marland et al. [1999] calculated that coal and oil burning in India contributed 70% and 23%, respectively, to fossil fuel derived CO₂ emissions. Cooke et al. [1999] have also pointed out the dominant contribution of coal and diesel to fossil fuel burning. These estimates do not refer to recent years, however, so we have estimated emissions for the year 2000 using projections from the Regional Air Pollution Information and Simulation Model for Asia (RAINS-Asia 8.0) [Downing et al., 1997]. These emissions reflect recent trends in fuel use and are determined from detailed energy-use data by fuel type and sector. Table 4 shows that direct coal combustion contributed 32% to total energy supply (22% from power generation and 10% from industrial manufacturing). Biofuels contributed 33%, and all oil products combined contributed 20%, the majority of which was diesel oil.

[51] The burning of fossil fuels and biofuels can release significant amounts of pollutants (gases and aerosols) into the atmosphere. The chemical analysis of the filter aerosol samples collected during INDOEX provided an opportunity to estimate the source strengths of these aerosols (e.g., fossil fuel or biomass burning), and to compare measurements with predictions based on emissions factors. In the following sections, we will present (1) emissions that have previously been calculated for species that are tracers either for biomass or fossil fuel burning, (2) results for our chemical analyses for the species of interest, and (3) a comparison between emissions and field measurement results.

3.6.2. Chemical Measurements

[64] In this section, ratios of the chemical tracers discussed above (i.e., BC, SO₄²⁻, K⁺, CO, and CH₃CN) are used to provide information in terms of the emission sources of the aerosols in the INDOEX region. Table 6 presents (1) a summary of the average BC/TC, BC/OC, K⁺/BC, SO₄²⁻/BC, SO₄²⁻/TC, BC/CO, and CH₃CN/BC ratios during INDOEX, (2) these ratios calculated (when possible) for burning of biofuel (wood and agricultural residues) using emission factors from Andreae and Merlet [2001], and (3) the same information for previous studies in urban and industrialized regions (from the mid Atlantic coast of USA, Europe, Korea, and Japan), and in regions influenced by biomass burning (in Africa and Brazil).

Table 6. Indicators for Biomass Burning or Fossil Fuel During INDOEX^a

Source	BC/TC	BC/OC	SO42−/BC	S/BCa	SO42−/TC	K+/BC	BC/CO	CH3CN/BC	References
INDOEX total	0.42 ± 0.09	0.77 ± 0.27	2.22 ± 0.73	0.74 ± 0.24	0.93 ± 0.43	0.15 ± 0.07	0.010 ± 0.005	0.05 ± 0.02	this study
INDOEX rCBL	0.38 ± 0.09	0.65 ± 0.22	1.47 ± 0.38	0.49 ± 0.13	0.67 ± 0.31	0.13 ± 0.05	0.012 ± 0.006	0.06 ± 0.01	this study
INDOEX MBL	0.43 ± 0.09	0.81 ± 0.28	2.35 ± 0.70	0.78 ± 0.23	1.00 ± 0.44	0.16 ± 0.07	0.009 ± 0.004	0.04 ± 0.02	this study
INDOEX Feb	0.46 ± 0.08	0.89 ± 0.20	2.24 ± 0.64	0.75 ± 0.21	0.98 ± 0.47	0.13 ± 0.05	0.012 ± 0.005	0.03 ± 0.01	this study
INDOEX March	0.36 ± 0.08	0.57 ± 0.19	2.18 ± 0.97	0.73 ± 0.32	0.80 ± 0.31	0.22 ± 0.07	0.008 ± 0.004	0.22 ± 0.19	this study
INDOEX Feb rCBL	0.38 ± 0.11	0.65 ± 0.27	1.27 ± 0.42	0.42 ± 0.14	0.65 ± 0.02	0.08 ± 0.01	0.017 ± 0.004	–	this study
INDOEX Feb MBL	0.47 ± 0.08	0.91 ± 0.25	2.32 ± 0.60	0.77 ± 0.20	1.05 ± 0.46	0.13 ± 0.05	0.011 ± 0.004	0.03 ± 0.01	this study
INDOEX March rCBL	0.39 ± 0.09	0.65 ± 0.21	1.61 ± 0.36	0.54 ± 0.12	0.70 ± 0.10	0.16 ± 0.02	0.009 ± 0.005	0.05 ± 0.01	this study
INDOEX March MBL	0.35 ± 0.07	0.55 ± 0.15	2.46 ± 1.08	0.82 ± 0.36	0.85 ± 0.37	0.24 ± 0.07	0.007 ± 0.003	0.19 ± 0.21	this study
Biofuel	0.11 ± 0.5	0.15 ± 0.6	0.70 ± 0.10	0.23 ± 0.40	0.081 ± 0.84	0.09 ± 0.3	0.008 ± 0.007	0.31 ± 0.16	Andreae and Merlet [2001]
Fossil fuelb	0.5 ± 0.05	1.0	–	–	–	0	–	0
ACE–2 (submicron), URB	0.38 ± 0.12	0.58 ± 0.22	15.2 ± 7.2	5.1 ± 2.4	5.5 ± 2.8	–	–	–	Novakov et al., 2000a
Seoul, Korea, URB	0.43 ± 0.05	0.76 ± 0.1	–	–	–	–	–	–	Kim et al., 1999
Sapporo, Japan, URB	0.49	0.97	1.1	0.37	0.55	0.04	–	–	Kaneyasu et al., 1995
EXPRESSO Africa, BB	0.20± 0.06	0.22	0.44	0.15	0.08	0.085	–	–	Ruellan et al., 1999
SCAR–B, Brazil, BB	0.10 ± 0.03	0.12 ± 0.03	0.28 ± 0.13	0.09 ± 0.04	0.03 ± 0.01	0.52 ± 0.11	–	–	Ferek et al., 1998

• a Literature Values are shown for comparison.
• a S/BC ratios are presented for comparison with ratios for emissions presented in the text. SO₄²⁻ and SO₂ values were converted into S (molecular interconversion).
• b assumed from diesel, household coal dominated [Kleeman et al., 2000].
• − = not available or not measured. INDOEX values are arithmetic averages. EXPRESSO (Experiment for Regional Sources and Sinks of Oxidants), Africa, savanna in the boundary layer, aircraft measurements. URB = urban and industrial pollution. BB = biomass burning.

3.6.2.1. Ratios of BC/TC and BC/OC

[65] The near constancy of BC/TC (0.42 ± 0.09) and BC/OC (0.77 ± 0.27) ratios for samples collected during the entire INDOEX period and the good correlation (r² = 0.86) between TC and BC (Figure 10) suggest that the INDOEX aerosol carbon is predominantly primary. If these species would have had similar source regions, the correlations could have been driven by meteorological factors. However, this does not seem the case because of the constancy of the ratios even when the BTs showed air masses of different origin (i.e., INDIA versus SEA versus WA). As shown in Table 6, the two ratios (BC/TC and BC/OC) are more similar to those measured in urban and industrial areas [Kaneyasu et al., 1995; Kim et al., 1999; Novakov et al., 2000a] than in areas affected by biomass burning [Ferek et al., 1998; Ruellan et al., 1999].

3.6.2.2. Ratio of SO₄²⁻/BC

[66] The aerosol SO₄²⁻ and BC measured during INDOEX are correlated, although r² is only 0.43 (Figure 11). The averaged SO₄²⁻/BC ratio was 2.22 ± 0.73. This ratio is much lower in areas strongly influenced by biomass burning (Table 6), suggesting that much of the aerosol pollution in the Indian Ocean represents a mixture of biomass and fossil fuel combustion.

[67] It is important to note here that international shipping has been found to be a major source of sulfur emissions in Asia, with the route from the Persian Gulf to the Indian Ocean having one of the highest sulfur emissions (and possibly particulate matter too) due to the traffic of oil tankers [Streets et al., 2000]. It is estimated that sulfur emissions in 1995 along this route were 53 Gg (S). This could be a possible reason for the higher SO₄²⁻/BC ratios observed at surface levels (cf. MBL samples with rCBL samples). This could also explain the results obtained on the R/V Ronald Brown during INDOEX, where an average SO₄²⁻/BC ratio of about 7.7 (about 3.5 times higher than our value) was found [Clarke et al., 2002].

3.6.2.3. Ratios of K⁺/BC

[68] The ratio of K⁺/BC should be higher for biomass burning aerosol than for fossil fuel aerosol, since K⁺ is emitted in high concentrations during most types of biomass burning [Andreae et al., 1983, 1998] and BC is released in lower concentrations. However, the K⁺/BC ratio is highly dependent on the type of biomass burned; therefore, its interpretation is not straightforward. A ratio of 0.15 ± 0.07 for K⁺/BC was obtained during INDOEX. Comparison with emission factors and interpretation of these results are discussed in the following 28.

3.6.2.4. Ratios of BC/CO and CH₃CN/BC

[69] BC/CO and CH₃CN/BC ratios were calculated for our samples using the CO and CH₃CN mass concentrations obtained by Campos et al. (personal communication, 2000) and Reiner et al. [2001], respectively. This resulted in an average ratio of 0.010 ± 0.005 for BC/CO and 0.05 ± 0.02 for CH₃CN/BC. There are no data available in the previous studies mentioned in Table 6 for the purpose of comparison. However, since CH₃CN background concentrations are in the range of 0.1 ppb or lower (D. Sprung, personal communication, 2000), the influence of biomass burning is evident from the fact that concentrations reached as high as 274 ppb in some of our samples.