a. Comparisons of airborne sun photometer and airborne in situ measurements of aerosol optical depths

Various layer AOD closure studies have been reported previously. Hegg et al. (1997) and Hartley et al. (2000) present the results of such studies based on measurements obtained in 1996 during the Tropical Aerosol Radiative Forcing Observational Experiment (TARFOX) field project, which took place in the same region as the present study. The layer AODs derived from the in situ measurements in those two studies were, on average, 14% ± 8% and 12% ± 5%, respectively, lower than the layer AODs measured by the airborne sun photometer. Remer et al. (1997) found that ground-based measurements of AOD made on the East Coast in July 1993 were nearly twice as large as AOD values derived from airborne in situ measurements; they attributed the difference to unmeasured aerosol layers aloft. In the Great Plains in 1997 and 1998, Kato et al. (2000) found that ground-based measurements of AOD compared well with values derived from airborne in situ measurements when the atmospheric RH was low, whereas the AOD derived from in situ measurements were ∼28% lower than those derived from ground-based measurements in higher RH cases. For data collected over the Atlantic Ocean (Canary Islands), Schmid et al. (2000) found that AODs from in situ airborne measurements generally did not agree with those measured by a sun photometer, possibly because particles were sampled through an inlet that was inadequate for sampling the large particles typical of marine and dusty air. In contrast, Magi et al. (2003) reported that in air masses in southern Africa dominated by small, biomass-burning aerosol and low RH, layer AODs derived from in situ measurements aboard the UW Convair-580, using the same instruments as those used to obtain the measurements discussed in the present paper, were only 4% ± 6% lower than those measured with the sun photometer aboard the aircraft.

In the present study, the airborne AATS-14 sun photometer provided remote sensing measurements of the column AOD. In a rapid tight climb (or descent) of the aircraft, the difference in column AOD between any two altitudes provides the AOD between those two points, which we call the “layer AOD.” We will refer to layer AODs measured by the sun photometer as τ_sun.

The layer AOD between heights z₁ and z₂(z₁ < z₂) due to particles is given by

View Expanded

where σ_ep(z) is the ambient light extinction coefficient due to particles at height z, and

σ_epzσ_spzσ_apz(2)

where σ_sp(z) and σ_ap(z) are, respectively, the ambient light scattering and the ambient light absorption coefficients due to particles at altitude z. However, before the measured values of aerosol light scattering and light absorption coefficients can be substituted into (2), and then used in (1) to compute a layer AOD from the in situ measurements for comparison with the ambient AOD measured by the sun photometer, they must be multiplied by appropriate hygroscopic growth factors to account for the difference between the measurements, which were made at a RH of ∼30%, and values at the ambient RH (the average ambient RH for the vertical profiles was about 60%). Hence, the AOD between heights z₁ and z₂ from the in situ measurements (τ_is), when adjusted to the ambient RH, is

View Expanded

where σ_spd is the dry light scattering coefficient due to particles, σ_apd the dry light absorption coefficient due to particles, and f_s and f_a are the RH-dependent factors by which the dry scattering and absorption coefficients, respectively, must be multiplied to adjust their measured values to the ambient RH.

Calculation of τ_is from (3) requires continuous measurements of σ_spd, σ_apd, f_s, and f_a between z₁ and z₂. The values of σ_spd and σ_apd were measured, and we used values of f_s as a function of RH (in percent) described by Kotchenruther et al. (1999); namely,

View Expanded

where a = 1.711 ± 0.044 and b = 3.41 ± 0.16 for airflow from the north or south, and a = 3.201 ± 0.022 and b = 3.78 ± 0.04 for airflow from the west. The values of f_a are unknown, but are likely to lie between 1 and f_s (Hegg et al. 1997; Redemann et al. 2001). Since σ_apd ≪ σ_spd, we assume that f_a = 1, which implies that σ_apd = σ_ap.

The common wavelength we use in comparing τ_sun with τ_is is 550 nm, at which all wavelength-dependent values are reported in this study. The nephelometer provided measurements of σ_spd at 550 nm. The PSAP provides measurements of σ_apd at 567 nm, but these were adjusted to 550 nm as described in section 2. The AATS-14 sun photometer provided the column AOD at 13 discrete wavelengths from 354 to 1558 nm; the closest wavelengths to 550 nm are 525 and 606 nm. To derive a column AOD at 550 nm from the AATS-14 measurements, a quadratic polynomial interpolation equation described by Schmid et al. (2003) was used. The uncertainty in τ_sun is mainly due to potential horizontal variability (Redemann et al. 2005) in the atmosphere during a vertical profile. Methods used to quantify this potential uncertainty are described in Redemann et al. (2003).

A sample vertical profile of in situ measurements of σ_sp, σ_apd, the aerosol single scattering albedo (ω₀), RH, and temperature (T) is shown in Fig. 1. The values of ω₀ were calculated at a wavelength of 550 nm from

View Expanded

The measurements shown in Fig. 1 were obtained on 17 July 2001, when the aerosol loading was the highest encountered during the CLAMS field study. Features of the vertical profiles shown in Fig. 1 include a distinct temperature inversion at about 0.4 km, and a mean RH of about 60% that clearly plays a role in the correlation between RH and the ambient light scattering coefficient σ_sp. The vertical profiles of the column AOD measured on 17 July 2001 by the sun photometer and derived from the in situ measurements are compared in Fig. 2, where the column AOD is defined by Eq. (1) with z₂ set to infinity. The column AOD from the in situ measurements is equal to that derived from (3) up to the maximum height of the aircraft plus the column AOD measured by the sun photometer at the maximum height of the aircraft. The column AOD measured by the sun photometer should (in theory) decrease monotonically with increasing altitude. However, this is not always the case in practice because of temporal fluctuations and spatial inhomogeneities in the atmosphere during the approximately 15 min it takes the aircraft to complete a vertical profile. The spatial inhomogeneity estimated using methods described by Redemann et al. (2003) suggested there was, on average, 27% ± 12% variability for every 100 km traveled horizontally in CLAMS. The vertical profiles discussed in this study covered an average horizontal distance of 20 ± 19 km, which resulted in small potential uncertainties in τ_sun (Table 3). The column AOD from the in situ measurements decreases monotonically with increasing altitude because it is an integrated quantity with no negative values.

b. Chemical composition of the aerosol

The total mass concentration of the aerosol with diameters <3 μm was measured gravimetrically, and the masses of sulfate, nitrate, and chloride ions were determined using methods described by Gao et al. (2003). The measured gravimetric and ion mass concentrations are listed in Table 4. Sulfate was the only ion consistently present in significant amounts. The masses of crustal elements and cations were not measured.

The total carbon (TC) component of the aerosol mass, defined as the sum of organic carbon (OC) and black carbon (BC), was measured using a thermal evolution method described by Kirchstetter et al. (2003). The results are listed in Table 4. The analysis did not separate OC and BC because the BC content was below the detection limit of the analysis method. However, BC was estimated to comprise, on average, less than 5% of the TC measured. Hegg et al. (1997) also reported that BC was a minor contributor to AOD in TARFOX.

The analysis of particles in CLAMS also provided the mass of oxalic acid (C₂H₂O₄), which is a dicarboxylic acid that originates from motor exhaust, dust, and soil sources (Kawamura and Kaplan 1987) and in secondary reactions between anthropogenic and natural organic compounds (e.g., Yokouchi and Ambe 1986; Sempere and Kawamura 1996). The oxalic acid mass concentration is implicitly included in the TC analysis, and is therefore not considered separately in the remainder of this study. As expected, the C₂H₂O₄ mass concentration was highly correlated (p < 0.05) with the TC values, with C₂H₂O₄ comprising, on average, 2% ± 1% of the TC mass concentration. The average oxalic acid mass concentration from Table 4 is 0.09 ± 0.04 μg m⁻³, which is lower than the average value of 0.5 ± 0.2 μg m⁻³ measured by Kawamura and Kaplan (1987) in Los Angeles. We attribute the difference to the lower contribution of motor exhaust off the U.S. Atlantic coast than in Los Angeles; the air parcel back trajectories discussed in section 4g may also have played a role in the low concentrations of oxalic acid in CLAMS.

c. Overall particle mass scattering efficiency and mass closure

The overall particle mass scattering efficiency (MSE) of the sampled aerosol was determined by a linear regression (e.g., Bevington and Robinson 1992) of the corrected nephelometer measurements of σ_spd to the measured gravimetric mass concentration (m_g). The results of this analysis are shown in Fig. 4, which yield an overall particle MSE of 2.4 ± 0.2 m² g⁻¹ (r = 0.95). The corresponding value for TARFOX was 2.8 ± 0.3 m² g⁻¹ (r = 0.88; Hegg et al. 1997). The error bars in Fig. 4 represent the standard deviation of σ_spd for the time the filter sample was collected and an uncertainty in m_g of ±5 μg propagated with the uncertainty in the sample flow volume (3% on average).

The mass closure analysis was performed for the sulfate and total carbon species (nitrate and chloride species do not improve the degree of closure). The measured mass concentrations of sulfate and total carbon listed in Table 4 account for only portions of their respective molecular forms in the atmosphere. We assumed the sulfate in the atmosphere existed in a one-to-one ratio with ammonium in the form of ammonium bisulfate NH₄HSO₄. Therefore, the measured sulfate mass concentrations were multiplied by 1.19 to convert them to mass concentrations of NH₄HSO₄. We assumed the total carbon mass concentration was half urban and half nonurban in origin and, using the results of Turpin and Lim (2001), arrived at a molecular conversion factor for carbon of 1.9 ± 0.3.

The mass closure results are shown in Fig. 5, where the error bars represent the uncertainty in m_g, as discussed above, and the propagated error for the mass concentration measurements. The error-weighted linear regression of m_g to the sum of the assumed atmospheric forms of sulfate and total carbon (Σ m_s) is

View Expanded

Hence, on average, m_g was 6% ± 6% greater than Σ m_s, with about 0.9 ± 0.5 μg m⁻³ unaccounted for. This implies that, for particles with diameters <3 μm, sulfate and total carbon accounted for an average of ∼94% of the total aerosol mass.

d. Species particle mass scattering efficiencies derived from multiple linear regression

The mass scattering efficiencies of individual chemical species can be estimated using the multiple linear regression (MLR) method described by White (1990). This method was used by Hegg et al. (1997) on the TARFOX data. The validity of this method was explored by Vasconcelos et al. (2001) in a comparison of theoretical model results with measurements obtained in southern California. They showed that interpretation of the MLR results requires the assumption that the aerosol be externally mixed, which implies that variations in the MSE of the various species are independent of the mass concentrations of other species.

Values of α_sulfate for a dry aerosol at a wavelength of 550 nm fall in the range ∼3–7 m² g⁻¹ (Charlson et al. 1999). Boucher and Anderson (1995) derived an average value of 3.4 ± 0.8 m² g⁻¹ for α_sulfate by varying parameters such as the particle size distribution and chemical form of atmospheric sulfate in Mie scattering model calculations. Values of α_sulfate ranged from 3.1– 5.1 m² g⁻¹ in studies reported between 1993–2000 (Haywood and Boucher 2000). Using the MLR method on data collected off the East Coast in the TARFOX field study, Hegg et al. (1997) derived a value of 3.2 ± 1.3 m² g⁻¹ for α_sulfate. White (1990) obtained a value for α_sulfate of 7.1 ± 0.5 m² g⁻¹ at a wavelength of 525 nm, which would be slightly lower for a wavelength of 550 nm.

Previous values of α_{total carbon} for a dry aerosol, derived from theoretical and measured particle size distributions inputted into Mie scattering models, range from ∼3–5 m² g⁻¹ (e.g., Shekar Reddy and Venkataraman 2000; Penner et al. 1998; Liousse et al. 1996). Zhang et al. (1994) derived an average value of 3.9 ± 1.5 m² g⁻¹ at a wavelength of 525 nm for data collected in the southwestern United States. Hegg et al. (1997) derived values of α_{total carbon} of 6.8 m² g⁻¹ using the MLR method on data collected off the East Coast in TARFOX. White (1990) used the MLR method to derive a value for α_{organic carbon} of 4.7 ± 0.5 m² g⁻¹ at a wavelength of 525 nm from a relatively large (47 data points) dataset collected mainly in the western United States. If we assume that in White's samples the mass of black carbon was 10% of the mass of organic carbon, the value for α_{total carbon} would have been 4.3 ± 0.5 m² g⁻¹. The values for α_{total carbon} given by Zhang et al., and derived from White's data, will decrease slightly when interpolated to a wavelength of 550 nm.

The values for α_sulfate and α_{total carbon} at a wavelength of 550 nm derived from the present study (6.0 ± 1.0 and 2.6 ± 0.9 m² g⁻¹, respectively) fall within a standard deviation of the range of values of earlier work. Our derived mean value for α_{total carbon} is generally lower than previously measured. This may be because the aerosol in CLAMS was internally mixed to the point where m_sulfate and m_{total carbon} were interdependent (Vasconcelos et al. 2001).

The results of MLR analyses can be affected by significant variations in the dry particle size distributions (Zhang et al. 1994). In general, the particle size distributions in this study were bimodal with a peak in the accumulation mode at diameter <1 μm and a second less distinct peak in the coarse mode at diameters between 1 and 4.5 μm. Total particle number concentrations averaged 800 ± 600 cm⁻³. The mean values of the geometric mean diameter, surface median diameter and geometric standard deviation of the data listed in Table 4 for the accumulation mode are d_g,acc = 0.19 ± 0.02 μm, d_smd,acc = 0.24 ± 0.03 μm and σ_g,acc = 1.40 ± 0.05, respectively. The corresponding mean values for the coarse mode are d_g,coa = 1.7 ± 0.1 μm, d_smd,coa = 2.5 ± 0.3 μm, and σ_g,coa = 1.8 ± 0.8. Hence, the accumulation mode (which accounts for the majority of the scattering) was stable to within ∼10%, which makes it unlikely that variations in the dry particle size distribution affected the MLR results reported here.

e. Chemical apportionment of the aerosol optical depth

The MLR results presented in the previous section can be applied to the individual vertical profiles that, together with the simultaneous measurements of m_sulfate and m_{total carbon} from the filter analysis (Table 4), can be used to estimate the contributions of sulfate and total carbon to the AOD. The values of m_x (the mass concentration of species x) were determined by averaging the individual m_x values for each UW flight to arrive at an average value (and standard deviation) for the vertical profile. Then, using the values of α_sulfate and α_{total carbon} determined in section 4.4, the contributions of sulfate and total carbon to the AOD were determined. For example, the fraction of light scattering contributed by sulfate, σ_sp-sulfate, is given by

View Expanded

The hygroscopic growth functions [Eq.(4)] were used to determine the difference between the AOD of a dry layer and a hydrated layer, which provides the contribution of condensed water to the layer AOD. The fraction of the layer AOD due to light absorption was determined from the PSAP measurements.

The contributions of sulfate, total carbon, condensed water and absorption to the layer AODs are listed in Table 5, and the percentage contributions of species to the total layer AOD are given in Table 6. Figure 6 shows these results in graphical form. The mean values of the contributions of sulfate, total carbon and condensed water to the layer AODs in CLAMS were similar, averaging 38% ± 8%, 26% ± 9%, and 32% ± 9%, respectively. Since the errors associated with the individual contributions vary, we calculated error-weighted averages; this yielded sulfate, total carbon, and condensed water contributions to the layer AODs of 39% ± 5%, 20% ± 5%, and 28% ± 7%, respectively. Clearly, sulfate was the largest contributor to the layer AOD in CLAMS. The contributions of absorption to the layer AODs were small (0.4%–4%) in all cases.

f. Single scattering albedo

The aerosol single scattering albedo (ω₀) is an important input to direct radiative forcing calculations. A sample profile of ω₀ is shown in Fig. 1. This particular case shows a much deeper polluted layer than was generally encountered in CLAMS (we define polluted layers as regions where σ_sp > 30 Mm⁻¹ in the vertical profiles listed in Table 2). Figure 7 shows the distribution of ω₀ values [calculated from Eq. (5)] in the polluted layer for all of the vertical profiles. The mean value of ω₀, based on 150 data points, is 0.96 ± 0.03.

Ground-based retrievals of ω₀ were also obtained by the Aerosol Robotic Network (AERONET) sun photometers (e.g., Dubovik et al. 2000) during the CLAMS field campaign from a site known as the Clouds and the Earth's Radiant Energy System (CERES) Ocean Validation Experiment (COVE; 36.9°N, 75.7°W). The vertical profiles in Table 2 were often spatially located close to COVE. The mean value of ω₀ at 550 nm from AERONET retrieval data (processed to remove clouds and manually quality assured) is 0.94 ± 0.03. Therefore, the mean value of ω₀ retrieved from AERONET agrees with mean value of ω₀ derived from our in situ airborne measurements (0.96 ± 0.03) to within one standard deviation.

On 17 July 2001, measurements were made from the UW aircraft and the COVE site that were both temporally (the aircraft vertical profile was from 1304–1337 UTC and the AERONET retrieval was at 1310 UTC) and spatially (the aircraft was ∼2.5 km from COVE) collocated. The mean value of ω₀ calculated from the airborne in situ measurements made in polluted layers during this vertical profile was 0.97 ± 0.02; the corresponding column-averaged value of ω₀ for accumulation mode particles retrieved from the AERONET data was 0.90 ± 0.03. Particle losses in the sampling system for the in situ instruments could have contributed to an underestimate of the absorbing component of the aerosol. Spatial variability may have played a role as well. Redemann et al. (2005) show that during CLAMS, the horizontal variability of AOD at midvisible wavelengths varied from 1% to 26% on scales less than 20 km. The values of ω₀ derived from the AERONET measurements are also subject to uncertainties at low solar zenith angles (54° at the time of the almucanter scan on 17 July) and for low AOD at midvisible wavelengths (Dubovik et al. 2000).

g. Comparison with TARFOX results

Since the measurements reported here were obtained in the same general location and time of year, and using the same techniques, as the TARFOX measurements reported by Hegg et al. (1997), it is of interest to compare them. The values of α_sulfate and α_{total carbon} derived from the TARFOX measurements were 3.2 ± 1.3 m² (g SO⁼₄)⁻¹ and 6.8 ± 1.1 m² (g C)⁻¹, compared to 6.0 ± 1.0 m² (g SO⁼₄)⁻¹ and 2.6 ± 0.9 m² (g C)⁻¹ derived here. The chemical apportionment reported here gives a higher mean sulfate contribution (38%) to the layer AOD at 550 nm than that obtained in TARFOX (19%). In the present study, the mean value of ω₀ at the ambient RH is 0.96 ± 0.03 (Fig. 7), and the mean value of ω₀ at 30% RH is 0.94 ± 0.04. For TARFOX, the mean value of ω₀ at ambient RH was ∼0.94, and the mean value of ω₀ at 30% RH was 0.90 ± 0.06.

The mean values of geometric mean diameter (d_g), the surface median diameter (d_smd) and geometric standard deviation (σ_g) for particle diameters <1 μm in this study (Table 4) are 0.19 ± 0.02 μm, 0.24 ± 0.03 μm, and 1.40 ± 0.05, respectively. The corresponding values for TARFOX were very similar: d_g,acc = 0.19 ± 0.02 μm, d_smd,acc = 0.24 ± 0.03 μm, and σ_g,acc = 1.40 ± 0.04. However, the mean value of the total particle number concentration in this study was 800 cm⁻³ while for TARFOX it was 1600 cm⁻³. The higher particle number concentrations clearly contributed to the higher average layer AOD in TARFOX.

Although the characteristics of the accumulation mode were similar in TARFOX and CLAMS, the sulfate contribution to the visible layer AOD was quite different. This implies that the differences in the values of α_sulfate and α_{total carbon} in TARFOX and CLAMS are due to differences in the size-resolved species concentrations of sulfate and total carbon. If this is the case, the sulfate (total carbon) concentrations in the accumulation mode in CLAMS are greater (less) than those in TARFOX. Unfortunately, since size-resolved species concentrations are not available for either TARFOX or CLAMS, we are unable to test this hypothesis.

Synoptic conditions could also account in part for the differences between the results presented here and those obtained in TARFOX. In the last two columns of Table 2 we summarize results from 72-h parcel back-trajectory analyses for the CLAMS case studies discussed here. The trajectories show that the air sampled off the East Coast in CLAMS generally had passed earlier over Canada, the Atlantic Ocean, or the Gulf of Mexico. For example, for UW flights 1870–1873, the air sampled above ∼200 m passed over Canada. For UW flights 1875–1882, the air sampled at all altitudes passed over either the Nova Scotia region or the Gulf of Mexico. For UW flight 1874, air up to ∼1.5 km passed over the immediate East Coast, while above 1.5 km it passed over the Great Lakes. Air sampled in UW flight 1874 was primarily ascending, but the air sampled in the other flights was generally subsiding. Figure 8 shows the results of 72-h back-trajectory analysis for UW flight 1874. In TARFOX the highest layer AODs occurred with airflows from the west and southwest, while the lowest layer AODs generally occurred with airflows from the northwest (Hegg et al. 1997). Airflows from the west and southwest pass over the large industrial and urban areas of the United States.

Since the airborne sampling and analysis techniques used in TARFOX and CLAMS were similar (and carried out by the same scientific team), we attribute the different aerosol characteristics encountered in these two field studies to the different origins of the sampled air.

h. Sensitivity of results to hygroscopic growth

Since the hygroscopic growth of the aerosol was not measured in this study, we assumed that the chemical composition and dry particle size distribution characteristics were similar to those measured in TARFOX and therefore used the parameterizations of hygroscopic growth described by Kotchenruther et al. (1999). Although the dry particle size distributions for particle diameters <1 μm (i.e., the accumulation mode) were similar in TARFOX and CLAMS, the dry sulfate contribution to the layer AOD in CLAMS was about twice that in TARFOX. The parcel back trajectories listed in Table 2 indicate that the aerosols sampled in CLAMS often came from the Canadian Atlantic coastal region or from sources in the Gulf of Mexico. As pointed out in section 4g, these back trajectories are different from those in TARFOX.

It could be that the results of Kotchenruther et al. (1999) underestimate the effect of RH for the aerosols measured in CLAMS. Assuming that Kotchenruther et al.'s results provide a lower bound for the RH growth in CLAMS, we can estimate an upper bound using Li et al.'s (2001) theoretical RH growth curve parameterizations for pure sulfate aerosols. We base our choice of the degree of acidity of the pure sulfate aerosol on a study by Brook et al. (1997), who found that sulfate aerosol acidity increased in maritime locations on the Canadian Atlantic coast compared to the generally neutralized sulfate aerosol encountered further inland. This was primarily due to the relative deficit of NH⁻₄ sources in maritime locations. Therefore, to estimate an upper bound for the RH growth in CLAMS, we will assume that the particles in air parcels that passed over the interior of North America were primarily neutralized (NH₄)₂SO₄, and that the particles in parcels that passed over the maritime regions near North America were primarily of acidic H₂SO₄.