2.1 ATR-42 measurement overview

This analysis focuses on flight missions conducted by the ATR-42 aircraft of SAFIRE (Service des Avions Français Instrumentés pour la Recherche en Environnement – the French aircraft service for environmental research) over the Gulf of Guinea and inland. A full description of flight patterns during DACCIWA is given in Flamant et al. (2018b). Here we present results from 15 flights focused on the characterization of anthropogenic pollution, dust and biomass burning plumes. The flight tracks are shown in Fig. 1 and a summary of flight information is provided in Table A1 in the Appendix. The sampling strategy generally consisted of two parts: first, vertical soundings were performed from 60 m up to 8 km above mean sea level (a.m.s.l.) to observe and identify interesting aerosol layers. Subsequently, the identified aerosol layers were probed with the in situ instruments by straight levelled runs (SLRs) at fixed flight altitudes.

Figure 1Tracks of the 15 flights analysed in this study. The colours indicate aircraft flight sampling layers dominated by biomass burning (green), mineral dust (orange), mixed dust–biomass burning (red), anthropogenic pollution (blue) and background particles (grey) from both vertical profiles (squares) and straight and level runs (SLRs; dots).

The ATR-42 aircraft was equipped with a wide variety of instrumentation performing gas and aerosol measurements. The measured meteorological parameters include temperature, dew point temperature, pressure, turbulence, relative humidity, as well as wind speed and direction. Gas-phase species were sampled through a rear-facing 1∕4 inch Teflon tube. Carbon monoxide (CO) was measured using ultra-violet and infrared analysers (PICARRO). The nitric oxide (NO) and nitrogen dioxide (NO₂) measurements were performed using an ozone chemiluminescence instrument (Thermo Environmental Instrument TEi42C with a Blue Light Converter for the NO₂ conversion). On-board aerosol instruments sampled ambient air via stainless steel tubing through the Community Aerosol Inlet (CAI). This is an isokinetic and isoaxial inlet with a 50 % sampling efficiency for particles with a diameter of 5 µm (Denjean et al., 2016).

The total number concentration of particles larger than 10 nm (N_tot) was measured by a butanol-based conductive cooling type condensation particle counter (CPC, model MARIE built by the University of Mainz; Mertes et al., 1995; Russell et al., 1996; Wiedensohler et al., 1997). The aerosol size distribution was measured using an ultra-high-sensitivity aerosol spectrometer (UHSAS, DMT, 0.04–1 µm), a custom-built scanning mobility sizer spectrometer (SMPS, 20–485 nm) and an optical particle counter (OPC, GRIMM model 1.109, 0.3–32 µm). Instrument calibration was performed with PSL nanospheres and oil particles for diameters from 90 nm to 20 µm. The SMPS data acquisition system failed after two-thirds of the campaign and could not be repaired. We found the UHSAS to show false counts in the diameters below 100 nm. Therefore, these channels were disregarded in the data analysis.

The particle extinction coefficient (σ_ext) at the wavelength of 530 nm was measured with a cavity-attenuated phase shift particle light extinction monitor (CAPS-PMex, Aerodyne Research). The particle scattering coefficients (σ_scat) at 450, 550 and 635 nm were measured using a TSI 3563 3-Wavelength Integrating Nephelometer and corrected for angular truncator error in the data inversion procedure using a Mie code (i.e. Sect. 2.3.1 and Fig. A1). The absorption coefficients (σ_abs) at 467, 520 and 660 nm were measured by a Radiance Research Particle Soot Absorption Photometer (PSAP). The PSAP measures changes in filter attenuation due to the collection of aerosol deposited on the filter, which were corrected for the scattering artifacts according to the Virkkula (2010) method. Prior to the campaign, the CAPS-PMex was evaluated against the combination of the integrating nephelometer and the PSAP. An instrument intercomparison was performed with purely scattering ammonium sulfate particles and with strongly absorbing black carbon (BC) particles. Both types of aerosol were generated by nebulizing a solution of the respective substances and size-selected using a DMA. For instrument intercomparison purposes, σ_ext from the combination of integrating nephelometer and PSAP was adjusted to that for 530 nm by using the scattering and absorption Ångstrom exponents (SAE and AAE, respectively). The instrument evaluation showed an excellent accuracy of the CAPS-PMex measurements by comparison to the integrating nephelometer and PSAP combination. The results were within the ±3 % uncertainty reported by Massoli et al. (2010) and Petzold et al. (2013) for the same instrument configuration.

2.2 Ancillary products

In order to determine the history of air masses prior to aircraft sampling, backward trajectories and satellite images were used. The trajectories were computed using the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) and the National Centers for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) data with 0.5^∘ horizontal resolution for sequences and times of interest. We compared the backward trajectory heights with information of fire burning times (e.g. MODIS Burnt Area Product) and dust release periods (e.g. Meteosat Second Generation (MSG) dust RGB composite images) to assess the aerosol source regions of the investigated air masses. The air masses represented by the trajectory are assumed to obtain their aerosol loading from source regions when the trajectory passes over regions with significant dust activation and/or fire activity at an altitude close to the surface. Trajectory calculations with slightly modified initial conditions with respect to the arrival time, location and altitude were performed to check the reliability of the location of source regions. Uncertainties in this approach, caused by unresolved vertical mixing processes and by general uncertainties of the trajectory calculations, are estimated to be in the range of 15 %–20 % of the trajectory distance (Stohl et al., 2002).

2.3 Data analysis

In the following, extensive aerosol parameters (concentrations, scattering, absorption and extinction coefficients) are converted to standard temperature and pressure (STP) using T=273 K and P=1013.25 hPa. The STP concentration data correspond to mixing ratios, which are independent of ambient pressure and temperature during the measurement. In the analysis, the data were averaged over sections of SLR with homogeneous aerosol conditions outside of clouds.

2.3.1 Derivation of aerosol microphysical and optical properties

Figure A1 and Table 1 show the iterative procedure and the equations used to calculate the aerosol microphysical and optical parameters as briefly explained below.

The particle number concentration in the coarse mode (N_coarse) was calculated by integrating the OPC size distributions over the range 1 to 5 µm. The signal-to-noise ratio of the OPC for particles in this size range was higher than 3, which makes the instrument well suited to quantifying variations in N_coarse. The number concentration of particles in the fine mode (N_fine) was obtained as the difference between total number concentration (N_tot particle diameter range above 5 nm) measured by the CPC and N_coarse.

Table 1Aerosol microphysical and optical properties derived in this work.

Download Print Version | Download XLSX

For optical calculations, the 3λ-σ_abs from the PSAP were adjusted at the three wavelengths measured by the integrating nephelometer using the AAE calculated from the 3λ measured σ_abs. Once σ_scat and σ_abs were obtained at the same wavelength, an optical closure study estimated the complex refractive index based on optical and size data. Optical calculations were performed using Mie theory, implying a sphericity assumption, because it facilitates a quantitative comparison with past data, mostly using this simplification and because most climate models assume spherical properties. The retrieval algorithm consists of iteratively varying the real part of the complex refractive index (n) from 1.33 to 1.60 and the imaginary part of the complex refractive index (k) from 0.000 to 0.080 in steps of resolution of 0.001. n and k were fixed when the difference between calculated values of σ_scat and σ_abs and measurements was below 1 %. Given that the size distribution measured by the UHSAS and the OPC depends on m, the optical-to-geometrical diameter conversion was recalculated at each iteration based on the assumed m. The resulting number size distributions from SMPS, UHSAS and OPC were parameterized by fitting four log-normal distributions and used as input values in the optical calculations. Once n and k were obtained at 3λ, we estimated the following optical parameters.

• SAE depends on the size of the particles. Generally, it is lower than 0 for aerosols dominated by coarse particles, such as dust aerosols, but it is higher than 0 for fine particles, such as anthropogenic pollution or biomass burning aerosol (Seinfeld and Pandis, 2006; Schuster et al., 2006).

• AAE provides information about the chemical composition of atmospheric aerosols. BC absorbs radiation across the whole solar spectrum with the same efficiency; thus, it is characterized by AAE values around 1. Conversely, mineral dust particles show strong light absorption in the blue to ultraviolet spectrum, leading to AAE values up to 3 (Kirchstetter et al., 2004; Petzold et al., 2009).

• SSA describes the relative importance of scattering and absorption for radiation. Thus, it indicates the potential of aerosols to cool or warm the lower troposphere.

• g describes the probability of radiation being scattered in a given direction. Values of g can range from −1 for entirely backscattered light to +1 for complete forward scattering light.

• MEE represents the total light extinction per unit mass concentration of aerosol. The estimates of MEE assume mass densities of 2.65 g cm³ for dust aerosol, 1.35 g cm³ for biomass burning aerosol, 1.7 g cm³ for anthropogenic aerosol and 1.49 g cm³ for background aerosol (Hess et al., 1998; Haywood et al., 2003a).

2.3.2 Classification of aerosol plumes

Data were screened in order to isolate plumes dominated by anthropogenic pollution from urban emissions, biomass burning and mineral dust particles, resulting in a total number of 19, 12 and 8 genuine plume interceptions, respectively, across the 15 flights. As shown in Fig. 2, identification of the plumes was based on a combination of CO and NO_x (sum of NO and NO₂) concentrations, as well as AAE and SAE, which have been shown to be good parameters for classifying aerosol types (Kirchstetter et al., 2004; Petzold et al., 2009). The classification was then compared with results from the back-trajectory analysis (Fig. 3) and satellite images described in Sect. 2.2.

Figure 2Absorption Ångström exponent (AAE) as a function of the ratio NO_x to CO. The markers are coloured according to the scattering Ångström exponent (SAE). Classification of mineral dust, biomass burning and urban pollution particles has been added to the figure.

Download

Figure 3Backward trajectories for the analysed aerosol layers. Trajectories date back 10 d for panel (a), 5 d for panels (b) and (c), and 1 d for panel (d).

The guidelines for classification are as follows.

• Anthropogenic pollution. SAE was beyond threshold 0, indicating a large number fraction of small particles in urban plumes and CO and NO_x concentrations 2 times higher than the background concentrations. During the DACCIWA campaign background CO and NO_x values were around 180 and 0.28 ppb, respectively. The trajectories show large differences in the flow patterns and source regions, with urban plumes originating from the polluted cities of Lomé, Accra and Abidjan. The aircraft sampling over land mostly followed the north-eastward direction (Fig. 3d).

• Biomass burning. The criteria are the same as for urban pollution plumes except that trajectories track these plumes back to active fire hotspots as observed by MODIS and the ratio NO_x to CO was set below 1. CO and NO_x are byproducts of combustion sources, but CO is preserved longer along the plume when compared with NO_x, which makes the ratio NO_x to CO a good indicator for distinguishing fresh anthropogenic pollution plumes from biomass burning plumes transported over long distances (Wang et al., 2002; Silva and Arellano, 2017). During this time of the year, most of the forest and grassland fires were located in central and southern Africa (Fig. 3a).

• Mineral dust. An AAE higher than 1 indicates a large mass fraction of mineral dust and a SAE below 0 indicates a high effective particle diameter. The source region of the dust-loaded air masses was located in the Saharan desert and in the Sahel (Fig. 3b).

• Dust and biomass burning mixing. Combining remote sensing observations and model simulations, Flamant et al. (2018a) identified a biomass burning plume mixed with mineral dust. This agrees well with the measured AAE of 1.2 and SAE of 0.3 observed in this layer. Menut et al. (2018) have shown that one of the transport pathways of biomass burning aerosols from central Africa was associated with northward advection towards Chad and then westward displacement linked to the African Easterly Jet. The plume originated from a broad active biomass burning area including Gabon, the Republic of Congo and the Democratic Republic of Congo and passed over areas with strong dust emissions further north within 1–3 d before being sampled by the aircraft (Fig. 3c).

• Background. We refer to background conditions as an atmospheric state in the boundary layer without the detectable influence of mineral dust, biomass burning or local anthropogenic sources. Most back trajectories originated from the marine atmosphere and coastal areas south of the sampling area.

3.1 Aerosol vertical distribution

Figure 4 shows a statistical analysis of N_fine, N_coarse and σ_ext derived from the in situ measurements of vertical profiles. The aerosol vertical structure is strongly related to the meteorological structure of the atmosphere (see Knippertz et al., 2017, for an overview of the DACCIWA field campaign). Therefore wind vector and potential temperature profiles acquired with the aircraft have been added to Fig. 4 as a function of the dominating aerosol composition, introduced in Fig. 1. The data from individual vertical profiles were merged into 200 m vertical bins from the surface to 6 km a.m.s.l. The profiles were calculated using only individual profiles obtained outside of clouds.

Figure 4Vertical layering of aerosols and meteorological variables for profiles for which aerosols dominated by biomass burning (green), dust (orange), mixed dust–biomass burning (red), anthropogenic pollution (blue) and background particles (black) were detected. The panels show profiles of (a) the extinction coefficient at 530 nm, (b) the particle number concentration in the range $\mathrm{0.005}<D_p<\mathrm{1}$ µm, (c) the particle number concentration in the range $\mathrm{1}<D_p<\mathrm{5}$ µm, (d) potential temperature, (e) the wind direction and (f) the wind speed. The coloured areas represent the 3rd, 25th, 75th and 97th percentiles of the data. The mixed dust–biomass burning plume is represented by a dot because it is derived from measurements during a SLR.

Download

The observed wind profiles highlight the presence of several distinct layers in the lower troposphere. For cases related to dust, urban pollution and background condition, we clearly observe the monsoon layer up to 1.5 km a.m.s.l., which is characterized by weak to moderate wind speeds (2 to 10 m s⁻¹, the latter corresponding to dust cases) and a flow from the south-west (220–250^∘). In all three air mass regimes, the monsoon layer is topped by a 500 to 700 m deep layer characterized by a sharp wind direction change (from south-westerly to easterly). Weak wind speeds (less than 5 m s⁻¹) are observed in urban pollution and background conditions, while higher wind speeds are observed in the dust cases. Above 2.5 km a.m.s.l., the wind speed increases in the urban pollution and dust cases and the wind remains easterly, indicating the presence of the African Easterly Jet with its core typically farther north over the Sahel (Fig. 8 in Knippertz et al., 2017, for the latitudinal variations of the African Easterly Jet during the DACCIWA field phase). The maximum easterlies are observed in the dust cases slightly below 3.5 km a.m.s.l. (>15 m s⁻¹). For the background cases, the wind above the shear layer shifts to north-westerly and remains weak (i.e. ∼5 m s⁻¹). Overall, the wind profile associated with the biomass burning cases is quite different from the other three cases, with a flow essentially from the south-south-west below 5 km a.m.s.l., higher wind speeds in the lower 2 km a.m.s.l. than above, and a secondary maximum of 7 m s⁻¹ at 4 km a.m.s.l.

The vertical distribution of aerosol particles was very inhomogeneous, both across separate research flights and between individual plumes encountered during different periods of the same flight. Measurements of aerosols within this analysis cover a broad geographic region, as shown in Fig. 1, which may explain some of the variability. SWA is subject to numerous anthropogenic emission sources (e.g. road traffic, heavy industries, open agriculture fires) coupled to biogenic emissions from the ocean and forests. These resulting large emissions are reflected in the high variability of σ_ext, N_fine and N_coarse in the lower troposphere over SWA. Below 2.5 km a.m.s.l., σ_ext showed a large heterogeneity, with values ranging from 35 to 188 Mm⁻¹ between the 3rd and 97th percentiles and a median value of 55 Mm⁻¹. The variability of σ_ext values was slightly enhanced near the surface and was correlated with N_fine and N_coarse, which ranged from 443 to 5250 cm⁻³ and from 0.15 to 1.6 cm⁻³, respectively. Maximum surface σ_ext was recorded in the anthropogenic pollution plume of Accra where high N_fine was sampled. The aerosol vertical profile is strongly modified during biomass burning and dust events. The dust plume extends from 2 to 5 km a.m.s.l. and is associated with transport from the dust sources in Chad and Sudan (see Fig. 3) with the mid-level easterly flow. The biomass burning plume extends from 1.5 to 5 km a.m.s.l. and is associated with transport from the south-west in a layer of enhanced wind speed just below 4 km a.m.s.l. as discussed above. Both layers showed enhanced σ_ext with median values of 68 Mm⁻¹ (p₀₃=12 Mm⁻¹; p₉₇=243 Mm⁻¹) in biomass burning plumes and 78 Mm⁻¹ (p₀₃=45 Mm⁻¹; p₉₇=109 Mm⁻¹) in dust plumes. As expected, the extinction profile was strongly correlated with N_fine for biomass burning layers and N_coarse for dust layers.

A prominent feature in the vertical profiles is the presence of fine particles up to 2.5 km a.m.s.l. outside of biomass burning or dust events. σ_ext, N_fine and N_coarse continuously decrease with altitude, most likely due to vertical mixing of local emissions from the surface to higher levels. Therefore, the regional transport of locally emitted aerosols was not limited to the surface, but occurred also at higher altitudes. Recently, numerical tracer experiments performed for the DACCIWA airborne campaign period have demonstrated that a combination of land–sea surface temperature gradients, orography-forced circulation and the diurnal cycle of the wind along the coastline favour the vertical dispersion of pollutants above the boundary layer during daytime (Deroubaix et al., 2019; Flamant et al., 2018a). Because of these complex atmospheric dynamics, aerosol layers transiting over the Gulf of Guinea in the free troposphere could be contaminated by background or urban pollution aerosols from the major coastal cities.

3.2 Aerosol size distribution

Figure 5a shows the range of variability of the number and volume size distributions measured during DACCIWA. These are extracted from the SLRs identified in Fig. 1. Figure 5b shows the same composite distribution normalized by CO concentration in order to account for differences in the amount of emissions from combustion sources.

Figure 5(a) Statistical analysis of number size distributions with coloured areas representing the 3rd, 25th, 75th and 97th percentiles of the data and (b) number size distributions normalized to CO for plumes dominated by biomass burning (green), dust (orange), mixed dust–biomass burning (red), anthropogenic pollution (blue) and background particles (black). In panel (b), the line thickness is scaled by the altitude of the aerosol plume.

Download

Considerable variability in the number concentration of the size distributions, up to approximately 2 orders of magnitude, was observed for a large fraction of the measured size range. The size distributions varied both for different aerosol types and for a given aerosol class. This reflects the relatively wide range of different conditions that were observed over the region, in terms of sources, aerosol loading, and lifetimes of plumes.

In particular for ultrafine particles with diameters below 100 nm, large differences were observed, with an increase as large as a factor of 50 in urban plumes, which reflects concentration increase from freshly formed particles. Interestingly elevated number concentrations of these small-diameter particles were also observed in some dust layers. Comparing the particle size distribution of the different dust plumes sampled during the field campaign, a variation as large as a factor of 20 in the number concentration of ultrafine particles is found (i.e. Fig. 4). Their contribution decreased with height, as reflected by the higher small particle number recorded in dust plumes below 2.5 km a.m.s.l. (Figs. 4b and 5b). As the composite urban size distributions showed a relatively similar ultrafine mode centered at 50 nm, dust layers have most likely significant contributions from anthropogenic pollution aerosol freshly emitted in SWA. The ultrafine mode was not observed in biomass burning size distributions, even though dust and biomass burning plumes were sampled in the same altitude range. We interpret this observation with dust plumes transported below 2.5 km a.m.s.l. that were sampled over the region of Savè (8^∘01^′ N, 2^∘29^′ E; Benin) near the identified urban air mass transported north-eastwards from Lomé and/or Accra and which may have collected significant fresh pollution on their way, whereas biomass burning plumes collected at the same altitude and sampled over the Ivory Coast south of the Abidjan pollution plumes may not have been affected by significant direct pollution (Fig. 1).

The accumulation mode was dominated by two modes centered at D_p,g ∼100 and 230 nm depending on the aerosol plume. The particle size distributions for biomass burning plumes were generally dominated by an accumulation mode centered at $D_{p,g}\sim \mathrm{230}$ nm. Despite the relatively wide range of sources and lifetimes of the biomass burning plumes sampled throughout the campaign (Fig. 3), the D_p,g in the accumulation mode showed little variation (D_p,g from 210 to 270 nm) between the plumes. Similarly, previous field studies found accumulation-mode mean diameters from 175 to 300 nm for aged biomass burning plumes, regardless of their age, transport time and source location (Capes et al., 2008; Janhäll et al., 2010; Weinzierl et al., 2011; Sakamoto et al., 2015; Carrico et al., 2016). The coagulation rate can be very high in biomass burning plumes and can shape the size distribution over a few hours (Sakamoto et al., 2016). It is worth noting that in the biomass burning and dust size distributions there is a persistent particle accumulation mode centered at ∼100 nm that exceeds the number of particles centered at 230 nm in some layers. This small mode is unlikely to be related to long-range transport of biomass burning and Saharan dust emissions, as it would be expected that particles in this size range would grow to larger particles through coagulation relatively quickly. As similar concentrated accumulation modes of particles have been observed in background plumes, it suggests the entrainment of background air from the boundary layer in dust and biomass burning plumes.

The number concentration of large super-micron particles was strongly enhanced in the mineral dust layers. The peak number concentration displayed a broad shape at D_p,g ∼ 1.8 µm, which is comparable to literature values of other long-range transported dust aerosols (Formenti et al., 2011a; Weinzierl et al., 2011; Ryder et al., 2013; Denjean et al., 2016; Liu et al., 2018). The super-micron mode of the dust plume is expected to be impacted by the mixing with other particles in case of an internal mixing, which should somewhat increase the particle size. The relatively homogeneous D_p,g in the coarse mode of dust reported here (D_p,g from 1.7 to 2.0 µm) suggests low internal mixing with other atmospheric species. Besides, the volume size distribution in urban plumes showed a significant presence (∼65 % of the total aerosol volume) of large particles with diameters of ∼1.5–2 µm, which were also observed in background conditions. We measured AAE in the range 0.7–1 in anthropogenic pollution plumes (Fig. 2), which suggests negligible contribution of mineral dust in these plumes. This coarse mode has most likely significant contributions from sea salt particles, as plumes arriving from the cities were transported at low altitude over the ocean (Fig. 3).

3.3 Aerosol optical properties

SSA is one of the most relevant intensive optical properties because it describes the relative strength of the aerosol scattering and absorption capacity and is a key input parameter in climate models (Solmon et al., 2008). Figure 6 shows the spectral SSA for the different SLRs considered in this study.

Figure 6(a) Statistical analysis of single scattering albedo at 450, 550 and 660 nm for plumes dominated by biomass burning (green), dust (orange), mixed dust–biomass burning (red), anthropogenic pollution (blue) and background particles (black). The boxes enclose the 25th and 75th percentiles, the whiskers represent the 5th and 95th percentiles and the horizontal bar represents the median. (b) Spectral SSA for the different individual plumes considered in this study. The mixed dust–biomass burning plume is represented by a dot because it is derived from measurements during only one SLR.

Download

The highest absorption (lowest SSA) at all three wavelengths was observed for biomass burning aerosols. SSA values ranged from 0.69 to 0.78 at 440 nm, 0.71 to 0.77 at 550 nm and 0.65 to 0.76 at 660 nm. This is on the low side of the range of values (0.73–0.93 at 550 nm) reported over West Africa during DABEX for biomass burning plumes mixed with variable proportion of mineral dust (Johnson et al., 2008). No clear tendency was found for the spectral dependences of SSA, which in some of the cases decreased with wavelength and in others were very similar to each other at all three wavelengths.

SSA values of anthropogenic pollution aerosols were generally intermediate in magnitude, with median values of 0.81 at 440 nm, 0.82 at 550 nm and 0.82 at 660 nm. Our data show that the value of SSA varied significantly for the different plumes. Some pollution aerosols absorb almost as strongly as biomass burning aerosols with SSA(550 nm) values as low as 0.72, whereas the highest SSA(550 nm) value observed was 0.86. In addition, the absorption properties of urban aerosol varied greatly between the sampled plumes for smoke of an apparently identical geographic origin. For example, we measured SSA(550 nm) values from 0.72 to 0.82 in the Accra pollution outflow. The variability in SSA values may be due to the possible contribution of emissions from different cities to the sampled pollution plumes (Deroubaix et al., 2019), thus having different combustion sources and chemical ages. The flat spectral dependence of SSA appears to be anomalous for anthropogenic pollution aerosols, as SSA has been shown to decrease with increasing wavelength for a range of different urban pollution plumes (Dubovick et al., 2002; Petzold et al., 2011; Di Biagio et al., 2016; Shin et al., 2019).

The magnitude of SSA increased at the three wavelengths when dust events occurred. Large variations in SSA were obtained, with values ranging from 0.76 to 0.92 at 440 nm, from 0.81 to 0.94 at 550 nm and from 0.81 to 0.97 at 660 nm. The measurement of SSA is highly dependent on the extent to which the coarse mode is measured behind the aerosol sampling inlet. Denjean et al. (2016) found that the absolute error associated with SSA, g and MEE of dust aerosols due to the CAI inlet is in the range covered by the measurement uncertainties. However, different aerosol inlet systems were used during previous field campaigns, which makes comparison of our results with previous measurements difficult. Overall, compared with the literature for transported dust, lower values were obtained in the present study for few cases. For example, Chen et al. (2011) reported SSA(550 nm) values of 0.97±0.02 during NAMMA (a part of AMMA operated by NASA) using an inlet with a comparable sampling efficiency. The lower values from DACCIWA reflect inherently more absorbing aerosols in some dust plumes. In contrast to fire plumes, the SSA of dust aerosol showed a clear increasing trend with wavelength. This behaviour is likely due to the domination of large particles in dust aerosol, which is in agreement with similar patterns observed in dust source regions (Dubovik et al., 2002). Moreover, an increase in SSA is observed with wavelength for mixed dust–smoke aerosol, suggesting that the aerosol particles were predominantly from dust, albeit mixed with a significant loading of biomass burning.

Table 2Single scattering albedo, mass extinction efficiency (in m² g⁻¹), asymmetry parameter and scattering Ångström exponent for the dominant aerosol classification.

Download Print Version | Download XLSX

As shown in Table 2, the observed variability of SSA reflects a large variability for MEE at 550 nm, which spans a wide range from 0.38 to 1.37, 1.45 to 1.92 and 1.24 to 4.83 m² g⁻¹ for dust and biomass burning for anthropogenic polluted aerosols, respectively. MEE is heavily influenced by the mass concentrations in the accumulation mode where the aerosol is optically more efficient in extinguishing radiation. We found MEE to be positively correlated with SAE (not shown), which was expected because of the dependence of MEE on particle size. In contrast, the values of g appear to differ only little between the sampled plumes for a given aerosol class. We found g in the range of 0.67–0.76 for dust, 0.65–0.68 for biomass burning and 0.59–0.64 for anthropogenic polluted aerosols at 550 nm. g values in dust plumes were high, which is expected due to the presence of coarse particles contributing to forward scattering.

This analysis includes sampled aerosols originating from different source regions and having undergone different aging and mixing processes, which could explain some of the variability. The impact of these factors on the magnitude and spectral dependence of optical parameters will be investigated in the following section.

4.1 Contribution of local anthropogenic pollution on aerosol absorption properties

Figure 7 shows the vertical distribution of SSA, SAE and the NO_x mixing ratio for the dominant aerosol classification. We exclusively consider measurements acquired during SLRs, since only during these phases was the whole set of aerosol optical properties measured. In dust plumes, if we exclude the case of mixing with biomass burning aerosol, SSAs were fairly constant above 2.5 km a.m.s.l., with values ranging between 0.90 and 0.93 at 550 nm, in agreement with values reported over dust source regions (Schladitz et al., 2009; Formenti et al., 2011b; Ryder et al., 2013, 2018). Despite the range of sources identified during DACCIWA, dust absorption properties do not seem to be clearly linked to particle origin or time of transport. Aerosols were more absorbing within the low-altitude dust plumes, with SSA values dropping to 0.81. SAE values exhibited simultaneously a sharp increase close to zero below 2.5 km a.m.s.l. This is consistent with a higher concentration of fine particles, though the value of SAE was still much lower than for pollution or background aerosol (i.e. where it is typically >0.2), which means that scattering was still dominated by larger particles. Based on the whole sets of observations, the strong variation in the light-absorption properties of dust-dominated aerosol over SWA could be attributed to the degree of mixing into the vertical column with either freshly emitted aerosols from urban/industrial sources or long-range transported biomass burning aerosol.

Figure 7Vertical distribution of (a) the single scattering albedo at 550 nm, (b) the scattering Ångstrom exponent and (c) the NO_x mixing ratio for the dominant aerosol classification. In panel (a), full circles represent SSA measurements and empty circles represent composite SSA calculated by deconvoluting size-distribution measurements in mixed dust layers and assuming an external mixing state.

Download

One of the critical factors in the calculation of aerosol direct and semi-direct radiative effects is the mixing state of the aerosols, which can significantly affect absorbing properties. There were no direct observational constraints available on this property during the DACCIWA airborne campaign. However, we investigated the probable aerosol mixing state by calculating composite SSA from the aerosol size distribution. On the basis of Fig. 5, dust size distributions showed only minor discrepancies in the mean and standard deviation of the coarse mode but significant differences in the balance between the fine and coarse modes, which suggests low internal mixing of dust with other atmospheric species. The size distributions of mixed dust pollution have been deconvoluted by weighting the size distributions of mineral dust and anthropogenic pollution aerosol averaged over the respective flights. This assumes that dust was externally mixed with the anthropogenic pollution particles and assumes a homogeneous size distribution for the dust and anthropogenic pollution aerosol throughout a flight. σ_scat and σ_abs were then calculated using Mie theory from each composite size distribution and the corresponding k and m. The refractive indices at 550 nm were assumed to be 1.52−0.002i and 1.60−0.040i for dust and anthropogenic pollution particles, respectively, which are the mean values deduced from the data inversion procedure (i.e. Sect. 2.3.1) throughout the campaign. The resulting σ_scat and σ_abs were used to calculate a composite SSA. A similar calculation was performed for the mixed dust–biomass burning case. Figure 7 shows a good agreement with the observations of SSA, implying that external mixing appears to be a reasonable assumption to compute aerosol direct and semi-direct radiative effects in these dust layers for modelling applications. This is consistent with the filter analysis performed during AMMA and SAMUM-2, which did not reveal any evidence of internal mixing in both mixed dust–biomass burning and dust–anthropogenic pollution layers (Chou et al., 2008; Lieke et al., 2011; Petzold et al., 2011).

SSA, SAE and NO_x of biomass burning plumes did not significantly vary with height from 2.2 to 3.8 km a.m.s.l. Moreover, the size distribution of biomass burning aerosols for the observed cases did not show a significant contribution of ultrafine particles (Fig. 5). These observations seem to indicate that the absorption properties of biomass burning plumes were not affected by direct pollution emissions.

In the boundary layer, the similar SSA and SAE in anthropogenic pollution and background plumes suggest that background aerosol may be rather called background pollution originating from a regional background source in the far field. Our analysis of the spectral dependence of SSA showed no apparent signature of anthropogenic pollution aerosols (see Sect. 3.3) despite a strong increase in aerosol number concentrations in air masses crossing urban centres (see Sect. 3.2). This can be explained by two factors. First, the majority of accumulation-mode particles were present in the background, while the large proportion of aerosols emitted from cities resided in the ultrafine-mode particles that have fewer scattering efficiencies (Fig. 5). Second, large amounts of absorbing aerosols in the background can minimize the impact of further increase in absorbing particles to the aerosol load. The high CO values (∼180 ppb) observed in background conditions further indicate a strong contribution of combustion emissions at the surface. Recent studies showed a large background of biomass burning transported from the Southern Hemisphere in SWA that dominated the aerosol chemical composition in the boundary layer (Menut et al., 2018; Haslett et al., 2019). The high absorbing properties (SSA ∼0.81 at 550 nm) and the presence of particles in both the accumulation and super-micron modes (i.e. Sect. 3.2) in background plumes are consistent with being a mixture of aged absorbing biomass burning and Atlantic marine aerosol. These results highlight that aerosol optical properties at the surface were dominated by the widespread biomass burning particles at regional scale.

4.2 Aging as a driver for absorption enhancement of biomass burning aerosol

The optical properties of aerosols are determined by either the aerosol chemical composition, the aerosol size distribution, or both. Changes in the size distribution of biomass burning aerosol due to coagulation and condensation have been shown to alter the SSA, as particles increase towards sizes for which scattering is more efficient (Laing et al., 2016). Variations in particle chemical composition, caused by source emissions and aging processes associated with gas-to-particle transformation and internal mixing, have been shown to change the SSA (Abel et al., 2003; Petzold et al., 2011).

In order to determine the contributions from size distribution and chemical composition to the variation of SSA in biomass burning plumes, SSA is presented as a function of SAE and k in Fig. 7a and b, respectively. k was iteratively varied to reproduce the experimental scattering and absorption coefficients, as described in Sect. 2.3.1. It appears that the variation of the size distribution (assessed via SAE in Fig. 8a) had minimal impact in determining the variability of SSA. Thus, the observations suggest that there was no effect of plume age on the size distribution, consistent with previous observations of size distribution in aged North American biomass burning plumes (Sakamoto et al., 2015; Carrico et al., 2016; Laing et al., 2016). Using a Lagrangian microphysical model, Sakamoto et al. (2015) have shown a rapid shift to larger sizes for biomass burning plumes within the first hours of aging. Less drastic but similarly rapid growth by coagulation was seen by Capes et al. (2008) in their box model. Given that the biomass burning plumes sampled during DACCIWA had more than 5 d in age, the quick size-distribution evolution within the early plume stages might explain the limited impact of the size distribution on the SSA.

Figure 8Contribution to single scattering albedo (a) from particle size (assessed via SAE) and (b) from composition (assessed via k) in biomass burning plumes.

Download

In contrast, Fig. 8b shows that there was a consistent decrease in SSA with increasing k, although there is some variability between the results from different plumes. The observed variability of SSA is reflected in a large variability of k, which is estimated to span the large range 0.048–0.080 at 550 nm. k depends both on the aerosol chemical composition and size distribution (Mita and Isono, 1980). Given that SSA was found to be independent of the aerosol size distribution (Fig. 8a), Fig. 8b suggests that SSA variability was strongly influenced by the variability in composition of biomass burning aerosol, implying a high contribution for light-absorbing particles. No clear tendency was found for the wavelength dependence of k, which in some of the cases increases with wavelength and in others decreases (not shown). In field observation, significant absorption and strong spectral dependence (values of AAE >1.7) in biomass burning plumes have been frequently attributed to the presence of brown carbon (BrC) (Kirchstetter et al., 2004; Sandradewi et al., 2008; Romonosky et al., 2019; Chakrabarty et al., 2010; Pokhrel et al., 2016). Contrary to the current understanding, our measurements show that the contribution of BrC to light absorption is negligible as AAE values ranged from 0.9 to 1.1 with a median value of 1.0. (Fig. 2b). Theoretically, fine-mode aerosol with absorption determined exclusively by BC would have an AAE equal to 1.0, since BC is expected to have a spectrally constant k (Bond et al., 2013). Therefore, the low SSA values observed in biomass burning plumes over SWA and the small spectral variation of k both suggest that BC is the dominant absorber in the visible and near-IR (infrared) wavelengths for these biomass burning aerosols.

Compared with past in situ measurements of aged biomass burning aerosol, SSA values over SWA (0.71–0.77 at 550 nm) are at the lower end of those reported worldwide (0.73–0.99 at 550 nm) (Magi et al., 2003; Reid et al., 2005; Johnson et al., 2008; Corr et al., 2012; Laing et al., 2016). This can be attributed in part to the high flaming versus smouldering conditions of African smoke producing more BC particles (Andreae and Merlet, 2001; Reid et al., 2005), which inherently have low SSA compared to other regions (Dubovick et al., 2002). However, SSA values over SWA are significantly lower than the range reported near emission sources in sub-Saharan Africa and over the south-eastern Atlantic, where values span over 0.84–0.90 at 550 nm (Haywood et al., 2003b; Pistone et al., 2019). Recent observations carried out on Ascension Island to the south-west of the DACCIWA region showed that smoke transported from central and southern African fires can be very light absorbing over the July–November burning season, but SSA values were still higher (0.80±0.02 at 530 nm; Zuidema et al., 2018) than those reported over SWA. A possible cause of the lower SSA in SWA is that Ascension Island is much closer to the local sources and the aerosol is therefore less aged.

Currently there are few field measurements of well-aged biomass burning emissions. Our knowledge of biomass burning aerosol primarily comes from laboratory experiments and near-field measurements taken within a few hours of a wildfire (Abel et al., 2003; Yokelson et al., 2009; Adler et al., 2011; Haywood et al., 2003b; Vakkari et al., 2014; Zhong and Jang, 2014; Forrister et al., 2015; Laing et al., 2016; Zuidema et al., 2018). With the exception of the study by Zuidema et al. (2018) over the south-eastern Atlantic, it is generally found that the aged biomass burning aerosol particles are less absorbing than freshly emitted aerosols due to a combination of condensation of secondary organic species and an additional increase in size by coagulation. This is in contrast to our results showing that SSA of biomass burning aerosols was significantly lower than directly after emission and that the evolution of SSA occurred a long time after emission.

There are three possible explanations for these results. First, one must consider sample bias. As regional smoke ages, it can be enriched by smoke from other fires that can smoulder for days, producing large quantities of non-absorbing particles, thereby increasing the mean SSA (Reid et al., 2005; Laing et al., 2016). However, during DACCIWA, biomass burning plumes were transported over the Atlantic Ocean and were probably less influenced by multiple fire emissions. Second, there is evidence that fresh BC particles become coated with sulfate and organic species as the plume ages in a manner that enhances their light absorption (Lack et al., 2012; Schwarz et al., 2008). Finally, organic particles produced during the combustion phase can be lost during the transport through photobleaching, volatilization and/or cloud-phase reactions (Clarke et al., 2007; Lewis et al., 2008; Forrister et al., 2015), which is consistent with the low SSA and AAE values we observed. Assessing whether these aging processes impact the chemical components and henceforth optical properties of transported biomass burning aerosol would need extensive investigation of aerosol chemical composition that will be carried out in a subsequent paper.