This paper investigates the properties of biomass burning aerosols over West Africa using data from the UK FAAM aircraft during the Dust and Biomass-burning Experiment (DABEX). Aged biomass burning aerosols were widespread across the region, often at altitudes up to 4 km. Fresh biomass burning aerosols were observed at low altitudes by flying through smoke plumes from agricultural fires. The aircraft measured aerosol size distributions, optical properties, and vertical distributions. Single scattering albedo varied from 0.73 to 0.93 (at 0.55 μm) in aerosol layers dominated by biomass burning aerosol. We attribute much of this variation to the variable proportion of mineral dust and biomass burning aerosol. We estimate the single scattering albedo of aged biomass burning aerosol to be around 0.81 with an instrumental uncertainty of ±0.05. External mixing, and possibly internal mixing, between the biomass burning aerosol and mineral dust presents an additional source of uncertainty in this estimate. The size distributions of biomass burning aerosols were dominated by particles with radii smaller than 0.35 μm. A 20% increase of count mean radius was observed when contrasting fresh and aged biomass burning aerosols, accompanied by changes in the shape of the size distribution. These changes suggest growth by coagulation and condensation. Extinction coefficients, asymmetry parameters, and Angstrom exponents are calculated from Mie theory, using the lognormal fits to the measured size distributions and assumed refractive indices.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Biomass burning aerosols influence the radiation budget of the earth and atmosphere by scattering and absorbing solar radiation [e.g., Haywood and Boucher, 2000; Abel et al., 2005; Myhre et al., 2003; Forster et al., 2007]. These aerosols can either warm or cool the climate, depending on the balance between scattering and absorption. In addition, scattering and absorption by aerosols reduces the solar radiative flux at the surface [e.g., Satheesh and Ramanathan, 2000]. This limits surface evaporation and surface heat fluxes, and may modify large-scale atmospheric circulations and precipitation patterns [e.g., Chung et al., 2002; Jones et al., 2007]. Biomass burning aerosols are generated mainly by agricultural burning in tropical regions [Andreae and Merlet, 2001]. Africa, the Amazon basin and southern Asia are major source regions.
 The optical properties of biomass burning aerosol vary regionally as different fires lead to differing aerosol chemical composition and physical properties [e.g., Eck et al., 2003; Reid et al., 2005a, 2005b]. The black carbon content is particularly important in governing absorption of solar radiation, but black carbon emissions vary with vegetation types and fire intensity [Reid and Hobbs, 1998]. Over time, biomass burning aerosols grow by coagulation, condensation and other gas-to-particle exchanges. This increases the scattering coefficient and single scattering albedo [Reid et al., 1998; Abel et al., 2003]. Aircraft measurements play an important role in understanding aerosols and developing model parameterizations. Recent aircraft measurements of biomass burning aerosols focused on South America [Kaufman et al., 1998; Chand et al., 2006] and southern Africa [Haywood et al., 2003a; Sinha et al., 2003]. The sub-Sahelian region of North Africa is another major global source of biomass burning aerosol, yet there have been few intensive observations of biomass burning aerosols over this region.
 The Dust and Biomass-burning Experiment (DABEX) was a major field campaign investigating the properties of biomass burning aerosols over West Africa during January-February 2006. The Facility for Airborne Atmospheric Measurement (FAAM) BAe 146 research aircraft was based at Niamey, Niger from 13 January to 3 February and made 13 research flights and measured aerosol and cloud properties, meteorological variables, and radiative fluxes. These aircraft measurements contributed to the African Monsoon Multidisciplinary Analysis (AMMA) dry season Special Observing Period (SOP-0) (J. M. Haywood et al., Overview of the Dust and Biomass-burning Experiment (DABEX), submitted to Journal of Geophysical Research, 2008). During the dry season, agricultural burning is intense across the sub-Sahelian region of Africa. This creates thick layers of biomass burning aerosol that extended for thousands of kilometers from east to west across the continent. Mineral dust aerosol was also present in high concentrations because of emissions from the Sahara desert and arid parts of the Sahel.
 The aim of DABEX was to measure biomass burning and dust aerosol properties for the development of model parameterizations and validate remote sensing techniques. Another objective was to study the variability in aerosol optical properties, such as differences between fresh and aged biomass burning aerosols. This paper presents observations of aerosol size distributions, single scattering albedos and vertical distributions for fresh and aged biomass burning aerosol. Mie calculations are also used to derive extinction coefficients, asymmetry parameters and Angstrom exponents. Furthermore, back trajectory analyses show the origins of observed dust and biomass burning aerosol layers. Mineral dust properties are investigated by Osborne et al. .
2. Aircraft Instrumentation
 The FAAM BAe 146 research aircraft contains a comprehensive suite of instruments [Osborne et al., 2007] measuring aerosol, cloud, chemical tracers, solar and terrestrial radiative fluxes and standard meteorological variables. Aerosol size distributions were measured by the Particle Measuring Systems (PMS) Passive Cavity Aerosol Spectrometer Probe 100X (PCASP), which sizes aerosols with radii in the range 0.05–1.5 μm. The TSI 3025A condensation nuclei counter (CNC) instrument measures submicron aerosol concentration by nucleating particles with butanol and counting them optically. This is able to detect all particles greater than 1.5nm in radius. A TSI 3076 nephelometer measures aerosol scattering coefficients at wavelengths of 0.45, 0.55, and 0.7 μm, which correspond to blue, green, and red light, respectively. A Particle Soot Absorption Photometer (PSAP) measures the aerosol absorption coefficient at 0.567 μm, inferred from attenuation of light through a filter.
2.1. Calibration of PCASP Instrument
 The PCASP is calibrated using latex spheres of known sizes that have a refractive index of 1.588 + 0.0i. However, absorbing particles such as dust and biomass burning aerosol scatter less efficiently than latex spheres causing the PCASP to underestimate their size. Therefore, correction have been applied to PCASP size bin radii, following Garvey and Pinnick  and Liu and Daum . These corrections are based on Mie calculations with an assumed refractive index of 1.54 + 0.045i for particles with uncorrected radii less than 0.3 μm (see section 7), and 1.53 + 0.0004i [see Osborne et al., 2008] for particles with uncorrected radii greater than 0.3 μm. The corrections increased size bin radii in the optically active region (0.1–0.3 μm) by 10–20%, and altered the PCASP size range from 0.05–1.5 μm to 0.051–1.60 μm.
2.2. Calibration of Nephelometer and PSAP Instruments
 The nephelometer data were calibrated using Anderson and Ogren  submicron corrections. These corrections compensate for the limited range of scattering angles (7–170°) detected by the nephelometer. The PSAP absorption coefficient was corrected for errors in filter exposure area, airflow rate, the influence of scattering by aerosols collected on the filter, and multiple scattering within the filter following Bond et al. . The Bond et al. corrections include a small adjustment to convert the absorption coefficient from 0.567 μm to 0.55 μm, the central wavelength of the nephelometer. This wavelength interpolation assumes that absorption varies as the reciprocal of wavelength, as in the work by Reid and Hobbs [1998, and references therein].
 The nephelometer and PSAP have Rosemount inlets that are inefficient at admitting particles with radii larger than about 1.5 μm [Haywood et al., 2003b]. Therefore, a correction is applied to account for scattering and absorption associated with particles greater than 1.5 μm based on comparisons with another PCASP mounted on the FAAM aircraft that had a wider size range (see Osborne et al.  for details). As shown later, the optical properties of the biomass burning aerosols were dominated by finer particles, in the region 0.051–0.35 μm, so these coarse particles corrections had a relatively small influence (5–10%) on scattering and absorption coefficients. The nephelometer and PCASP heat air in the instrument, which may lead to partial or complete evaporation of any water present within the aerosols. However, since the ambient relative humidity averaged about 40%, and did not exceed 70% on any of the runs, it is unlikely that the aerosols contained significant water.
3. Aircraft Measurement Strategy
 All flights started and ended at Niamey International Airport, Niger. Six flights went to the south to measure biomass burning emissions from the belt of fires around 5–12°N (Figure 1). Four flights near Niamey investigated aged biomass burning layers and mixing between dust and aged biomass burning aerosol. Additionally, three flights investigated plumes of Saharan dust over northeast Niger [Osborne et al., 2008]. Flights involved profiles to investigate the vertical distribution of aerosols, and straight and level runs to sample aerosols at different altitudes (Haywood et al., submitted manuscript, 2008).
3.1. Sampling of Fresh Biomass Burning Aerosol From Individual Smoke Plumes
 The properties of fresh biomass burning aerosols were investigated on two flights (B162 on 24 January, and B166 on 1 February). These focused on a region 9–11°N, 3–6°E, where agricultural burning was particularly intense. Visual observations suggested that most biomass fires were small (burn areas of 1–10 hectares) and burnt a mixture of grasses, scrub vegetation and agricultural debris. Fires generally involved a mixture of flaming and smouldering combustion. Smoke plumes were sampled by flying over a large number of small fires, approximately 300 m above ground level. Locations of runs sampling fresh smoke plumes are shown in Figure 2.
Figure 3 shows an hour of data from flight B166 (1 February, 1330–1430 UTC, 10°N, 5–6°E). This is an example from the 5 h of data collected during flights B162 and B166 targeting individual fires. Smoke plume penetrations are best identified by peaks in the PCASP aerosol concentration since fires create a very high local concentrations of particle number. The passage of the aircraft over the visually observed fires was noted by sharp increases in aerosol number concentration from the PCASP and CNC concentration and peaks in CO concentration. The correspondence between PCASP and CO concentrations are shown in Figure 3b, which supports the assumption that these are associated with biomass burning emissions. The peaks in PCASP concentration (Figure 3a) have temporal scales of 10–50 s, which corresponds to horizontal scales of 1–5 km (since the aircraft travels at approximately 100 ms−1), which shows that the aerosol is not of a highly dispersed nature, as would be expected from aged transported aerosol. As shown in section 5.3 the CNC concentrations was extremely high within the fresh aerosol samples, compared to the CO concentration, which is indicative of a fresh aerosol population with many fine particles. The nephelometer scattering coefficient and PSAP absorption coefficients do not correlate as well with smoke plume penetrations because the smoke particles are smaller and consequently have a lower scattering cross-sectional area than the background dust.
 To characterize the properties of fresh biomass burning aerosols it was necessary to separate smoke plume penetrations from the background aerosol. Since high aerosol number provided the best indicator of smoke plumes penetrations, smoke plumes are defined by PCASP concentrations over 5000 cm−3 as in the works by Haywood et al. [2003a] and Abel et al. . This threshold is illustrated in Figure 3a. Additionally, only plumes with 30 s or more of continuous data were included. This limited errors associated with instrument response time delays. Fresh biomass burning smoke plumes make up 16 min of data from the 5-h time series, which, while not exhaustive, should be a sufficient sample size to minimize sampling error. It is difficult to estimate the typical ages of the aerosol in the data selected as “smoke plumes”, but the evidence presented above clearly distinguishes it as being relatively young compared to the extensive and more uniform layers observed at higher altitudes.
3.2. Runs Dominated by Aged Biomass Burning Aerosol
 Layers of aged biomass burning aerosol were observed on every flight during DABEX, and were widespread across the West African region. However, the biomass burning aerosol was always mixed, to some degree, with mineral dust. The dust was also noted by a steady increase in the volume distribution for coarse particles with radii greater than 0.35 μm (as shown later in Figure 8 from section 5.1). This mixing with dust tended to give much lower Angstrom exponents than expected for biomass burning aerosols. Figure 4 shows the average Angstrom exponent from the nephelometer (using the 0.45 and 0.7 μm channels) for each aircraft run during DABEX, plotted against height. Angstrom exponents vary from -0.1 to 1.5 and generally increase with height as the proportion of dust declined at higher altitudes. The lowest values coincide with runs that have been confirmed by filter analyses as purely dust cases [Formenti et al., 2008]. The higher values of Angstrom exponent coincide with runs where the size distribution, and CO concentration are indicative of biomass burning aerosol.
 Therefore, an Angstrom exponent threshold of 1.0 has been used to categorize runs in which the aerosol scattering (and hence optical properties) are expected to be dominated by biomass burning aerosol, rather than being dominated by dust. This threshold is arbitrary and has been set to 1.0 to ensure a reasonable sample size of aged aerosol runs and exclude runs where dust has a strong influence on aerosol optical properties. As a further measure, a threshold aerosol extinction coefficient of 0.03 km−1 is used to exclude dilute aerosol layers. Runs shorter than 5 min are also excluded to minimize sampling error. Following these criteria, 16 runs are selected to characterize aged biomass burning aerosols. These are gathered from 7 different flights and make up approximately 3.5 h of data. Figure 2 shows the locations of these runs.
4. Vertical Distribution of Aerosols
 The vertical distribution of aerosol was of particular interest during DABEX as distinctive layers of biomass burning and dust were often found at different altitudes. This vertical structure determined how much mixing took place between the biomass burning and dust. The vertical distribution also has an influence on the direct radiative forcing of the aerosols, as shown by Johnson et al. .
4.1. Elevated Layers of Biomass Burning Aerosol: Near Niamey
 Elevated layers of aged biomass burning aerosols were found near Niamey on every flight during DABEX, usually at altitudes of 1–4 km. An example is shown in Figure 5, taken from a profile near Niamey airport on 19 January 2006 (flight B159). The profile shows a deep layer of biomass burning aerosol from 1 to 3.5 km with a weak layer of dust aerosol beneath. The nephelometer (Figure 5a) shows a preference for scattering at shorter wavelengths (blue and green) in the upper layer, which indicates high concentrations of fine particles (most likely biomass burning aerosol). This is accompanied by a high CO concentration in the upper layer (Figure 5b), a further indication of combustion-related (i.e., biomass burning) emissions. The PCASP volume is shown in Figure 5c and is also segregated into fine (radii 0.051–0.35 μm) and coarse (0.35–1.6 μm) contributions. The fine volume peaks in the upper layer and correlates well with the CO concentration suggesting that biomass burning aerosol dominates the fine volume. The fine volume also correlates well with increased nephelometer scattering in the upper layer, and increased wavelength dependence of the scattering. The coarse volume is more constant with height and is assumed to be dominated by mineral dust aerosol. The coarse mode dominates the aerosol volume in the lower aerosol layer (0–1 km) (Figure 5c), which is consistent with the very low wavelength dependence in the scattering coefficient measured by the nephelometer (Figure 5a). There is also a small elevation of CO and fine aerosol volume in the lower layer, which is evidence of a small quantity of combustion-related fine particles that have mixed with the dust. These could be biomass burning aerosols from local fires, or industrial aerosols from Niamey.
 The elevated biomass burning aerosol layer (1.5–3 km) corresponds to a weakly stratified layer, capped by a slight temperature inversion at 3–3.5 km. This is probably the residual of a deep continental boundary layer that originated several hundred kilometers to the south and has been elevated and transported north by the mean atmospheric circulation. The residual continental boundary layer has been undercut by cooler, drier air, probably originating from the Sahara, to the north. The lower layer is thermodynamically well mixed from the surface to 600 m typical of a developing daytime boundary layer; the profile was made at 0930 local time. A stable layer from 600 to 1000 m prevents mixing between the upper and lower layers, preserving a relative clear slot at 1 km.
 The Met Office dispersion model [Manning et al., 2003] was used to investigate the origin of the low level dust and elevated biomass burning aerosol. Sixty-eight thousand particles were released in the model at 0900 UTC, over Niamey (13.5°N, 2.5°E) at 0.5 km and 2.5 km, corresponding to the locations of the dust and aged biomass burning aerosol layers observed in the profile of Figure 5. Each particle's trajectory was traced back in time for 5 days using meteorological analyses from the Met Office operational Numerical Weather Prediction model. Each particle's trajectory was randomly perturbed at each time step to reflect the influence of turbulent motion, resulting in a ensemble of trajectories. Figure 6 highlights regions where back trajectories are found within lowest 500 m of the atmosphere. This illustrates possible source regions of the aerosol since smoke or dust is initially emitted into the lowest levels of the atmosphere. The aged biomass burning aerosol observed at 2.5 km originated from the southeast suggesting sources in northern Nigeria. The dust layer observed at 0.5 km originated from the north indicating sources in the central Sahara.
4.2. Dust and Fresh Smoke at Low Levels: South of Niamey
 Over the intense biomass burning regions to the south of Niamey (6–12°N) the daytime boundary layer was often 2–3 km deep, and typically contained a mixture of dust, and fresh or aged biomass burning aerosol. Figure 7 shows a profile taken over western Nigeria (approximately 10°N, 5°E) during one of the fresh smoke plume flights (B166 on 1 February 2006 at 1430 local time). The nephelometer indicates a mixture of biomass burning and dust as the wavelength dependency of scattering is not as strong as expected from pure biomass burning aerosol. The nephelometer scattering, CO concentration and PCASP volume are all exceptionally high (Figures 7a–7c) in the boundary layer (below 2 km), indicating high levels of biomass burning aerosol and dust. The CO concentration is more variable than the nephelometer scattering or aerosol volume (Figures 7a–7c). These peaks probably correspond to fresh smoke plumes intercepted during the profile, since many fires and smoke plumes were observed visually in that region during the flight. Because the fresh smoke is composed of fine optically inefficient particles (small scattering per particle), the interception of fresh smoke plumes are less prominent in the nephelometer profile or the PCASP volume. There is however, some correspondence between the CO peaks at 0.6 km and 1.4 km and the fine aerosol volume (Figures 7a and 7c).
 The drop off in aerosol concentrations above 2 km corresponds to a slight temperature inversion at 2 km accompanied by a 5–10 K increase in dew point temperature (i.e., increased humidity) around 2–2.5 km (Figure 7d). The aerosol above 2 km may be the residual of boundary layers from previous days. Alternatively, vigorous smoke plumes may have penetrated through the top of the boundary layer, because of additional buoyancy associated with heat from fires.
5. Aerosol Size Distributions
Figures 8–11 show aerosol size distributions from the PCASP instrument as a function of corrected particle radius (see sections 2.1 and 2.2 for a description of corrections related to refractive index). In this analysis the corrected PCASP data covers the size range 0.051–1.6 μm, as opposed to the manufacturer's stated range of 0.05–1.5 μm for uncorrected radii. The contribution to extinction, as a function of particle size, has also been calculated for each PCASP size bin by performing Mie calculations with the midbin radii and refractive indices as described in sections 2.1 and 2.2.
Figure 8 shows the size distributions obtained in fresh smoke plumes (following the “smoke plume” criteria in section 3.1). This is compared against the background aerosol size distributions from the smoke plume flights (details also given in section 3.1) and a mineral dust size distribution from two flights (B160 and B161) investigating dust over northeast Niger [see Osborne et al., 2008]. The number concentration plot (Figure 8a is not normalized and reflects absolute differences in particle concentrations that were observed. This emphasizes the very high concentration of fine particles associated with smoke plumes. The concentration of coarse particles (0.35–1.6 μm) is also slightly higher in smoke plumes compared to the background, but is not as high as in the mineral dust case. We assume that most of the coarse particles are mineral dust, although some may be fly ash. It is possible that convection associated with fires helps to lift dust from near the surface giving higher dust concentrations in dust plumes.
 The size distributions are shown again in Figure 8b as volume plots and are normalized so that the total volume is equal over the coarse size range (0.35–1.6 μm). This normalization emphasizes the consistency in the shape of the size distribution over the coarse size range. In volume space it is also clear that the distributions have two distinct modes: a fine mode that dominates for radii 0.051–0.35 μm and a coarse mode that dominates for radii 0.35–1.6 μm. The fine mode is assumed to be predominantly biomass burning aerosol and the coarse aerosol component is assumed to predominantly mineral dust. The consistency in the shape of the coarse mode may suggest that the mineral dust is from a common source, for example a similar region of the Sahara dessert.
 On the basis of Figure 8 the smoke plume size distribution has been separated into two components: one for the dust, one for the fresh biomass burning aerosol. These components are assumed to be externally mixed together in the smoke plume [Formenti et al., 2008]. The dust component is assumed to have the same size distribution as the mineral dust case shown in Figure 8. This assumes that dust size distributions do not vary between flights B162, B166 (measuring dust and fresh biomass burning aerosol mixtures) and B160 and B161 (measuring mineral dust only). This assumption seems reasonable over the coarse size range, as shown in Figure 8b, although it is impossible to tell whether fine-particle dust concentrations vary between the mineral dust case and the dust contained within the smoke plumes. However, since the concentration of fine particles in the smoke plume far outweighs that observed in the background aerosol we assume that the fine particles are strongly dominated by biomass burning emissions. Therefore, the uncertainty associated with variable proportions of fine to coarse dust is assumed to be acceptably small for the purposes of this analysis. The fresh biomass burning aerosol component is simply derived by subtracting the dust component from observed size distribution until there are no particles with radii greater than 0.35 μm. This assumes that biomass burning aerosols do not contribute significantly to the size distribution beyond 0.35 μm. Past observations [Reid and Hobbs, 1998; Haywood et al., 2003a] suggest this is a reasonable assumption.
 The size distribution of the fresh biomass burning aerosol component is shown in Figure 9, and is compared with PCASP data from SAFARI-2000. The SAFARI data is from a large biomass burning smoke plume, observed by the Met Office C-130 aircraft over Namibia on 13 September 2000 [Abel et al., 2003; Haywood et al., 2003a]. The dust concentration during SAFARI was an order of magnitude lower than in DABEX, so it was not necessary to separate SAFARI data into biomass burning aerosol and dust components, as we have done here for the DABEX data. The size distributions in Figure 9 have been truncated so that they only cover the size range 0.05 to 0.35 μm, where biomass burning aerosols are dominant.
 The number concentration is dominated by particles at the smallest sizes within the PCASP size range (e.g., 0.051–0.1 μm, Figure 9a) whereas most of the volume and extinction comes from particles in the size range 0.07–0.25 μm (Figures 9b and 9c). The extinction is more closely related to the volume than the number concentration. The shapes of the DABEX and SAFARI-2000 size distributions look very similar (Figure 9), even though the SAFARI-2000 and DABEX experiments took place in different geographic regions. The shape of the DABEX fresh biomass burning component was not sensitive to the smoke plume identification criteria defined in section 3.1. This consistency is encouraging for model parameterization development.
 For modeling applications, it is more convenient to represent aerosol size distributions by formulaic expressions, such as a lognormal distribution, rather than using instrumental data. A lognormal distribution has been fitted to the DABEX fresh biomass burning aerosol component, and is shown in Figure 9, compared against the instrumental data. The lognormal has a geometric mean radius of 0.08 μm and a standard deviation of 1.4. (as listed in Table 1). The number, volume, and extinction distributions are captured reasonably well by this model fit, although the PCASP curves are not somewhat irregular in shape and the smooth lognormal cannot capture all these features. Since the model fits will be used to derive aerosol optical properties, it is most important that the extinction peak in the model fit coincides with the extinction peak in the instrumental data (Figure 9c).
Table 1. Lognormal Distribution Parameters for Fresh and Aged Biomass Burning Aerosol Size Distributions
 The mean size distribution observed for runs dominated by aged biomass burning aerosols is shown in Figure 10 and is contrasted against the size distributions from two individual runs, one that appeared to have fresher aerosol (“less aged”), and another that appeared to have more aged aerosol (“v. aged”). These show a large degree of variability in the slope of the number distribution (Figure 10a) over the size range 0.051–0.2 μm. This is associated with a reduction of fine particles (radius 0.05–0.1 μm) and the growth of intermediate-sized particles (radius 0.1–0.2 μm). In the volume plot (Figure 10b), these changes are exhibited by a shift in the fine mode toward larger radii. There is also a negative skewing of the fine mode peak with the “aged” and “v. aged” cases. We assume that these changes are caused by coagulation and condensation during the ageing process. Coarse particle concentrations (radii 0.35–1.6 μm) also vary, but we attribute this to variability in the proportion of mineral dust.
 The dust component of the size distribution was removed following the same method as outlined for fresh aerosols in the previous section. Again, this assumes that the dust is externally mixed with the biomass burning aerosol and assumes a size distribution for the dust aerosol based on flights over northeast Niger. The size distribution component associated with aged biomass burning aerosol is shown in Figure 11 and compared against aged biomass burning aerosol data from SAFARI-2000 [Haywood et al., 2003a]. The DABEX and SAFARI-2000 number size distribution are both fairly flat over the size range 0.05 to 0.15 μm, then drop sharply from 0.2 to 0.35 μm. The volume distributions have a similar shape but the peak in the volume occurs at a slightly larger size of around 0.2 μm in DABEX, compared to 0.15 μm in SAFARI-2000 (see Figure 11b). This difference could be due to differences in the rates of coagulation and condensation, or differences in PCASP performance between DABEX and SAFARI-2000. The volume and extinction peaks occur at slightly larger sizes than in the fresh aerosol cases because of the growth of particles at around 0.15 μm.
Figure 11 also shows the lognormal model that has been fitted to the DABEX aged biomass burning data. This has two modes, as illustrated by the dashed lines in Figure 11. The first mode has the same mean radius (0.08 μm) and standard deviation (1.4) as chosen for the fresh smoke plume data, and could be interpreted as representing remaining fresh aerosols that have not grown. The second mode has a larger mean radius (0.14 μm) but a lower standard deviation (1.25) (as listed in Table 1) and could be interpreted as representing particles that have grown as a result of ageing. Although this interpretation does not represent the continuous nature of the ageing process, the categorization of a fresh and aged mode may be useful in some modeling applications.
 The lognormal model fit captures the DABEX observed size distributions reasonably well in number, volume, and extinction space. The model fit should therefore be provide a reasonably accurate basis for the derivation of aerosol optical properties from Mie theory. The second mode of the lognormal model will be of particular importance to aerosol optical properties, since the extinction peak between 0.13 and 0.2 μm is dominated by the contribution from the second mode (Figure 11c).
5.3. Evidence for Coagulation Between Fresh and Aged Aerosol Samples
Table 2 compares count mean radii and ratios between PCASP concentration, PCASP fine volume, CNC concentration, and excess CO in fresh and aged aerosol samples. The count mean radius increases by about 20% between the fresh and aged aerosol indicating a moderate growth of particles during the aging processes. This could be due to either condensation or coagulation, or combination of both. The count mean radii are 20–40% larger than the values reported from the Smoke Clouds and Radiation–Brazil (SCAR-B) experiment in South America [see Reid et al., 1998]. This may be due to the influence of coarse dust particles that were more abundant in DABEX than SCAR-B. Volume mean radii have not been shown since the mineral dust dominated the aerosol volume, whereas the biomass burning aerosol dominated the aerosol number (count).
Table 2. Count Mean Radii and Ratios Between PCASP Concentration, PCASP Fine Volume, CNC Concentration, and CO Excess Concentration (ΔCO = CO - 125 ppbv)a
All values are averages taken over all data categorized as “fresh” or “aged” according to selection criterion in sections 3.1 and 3.2. Units are μm for count mean radii, cm−3 for PCASP and CNC concentrations, and μm3 for PCASP fine volume (f.v.).
Count mean radius
 The excess CO concentration (ΔCO) used here is the CO concentration minus a “background” value of 125 ppbv, which was determined from unpolluted air samples measured during DABEX. ΔCO provides a useful tracer of biomass burning emissions since it is not readily removed in the atmosphere. Particles however, are readily lost through coagulation and deposition processes. The ratio of PCASP concentration to ΔCO decreases by a factor of 3.7 between the fresh and aged aerosol samples and the ratio of CNC concentration to CO decreases by a factor of 7.4 between the fresh and aged aerosol. The loss rate is higher for the CNC concentration than the PCASP because the CNC is able to detect very fine particles, down to 1.5 nm radius, that coagulate more rapidly than those in the PCASP size range (see section 2). This is further highlighted by the decrease in the ratio between CNC concentration and PCASP concentration between fresh and aged aerosol data.
Table 2 also shows an almost threefold decrease in the ratio of PCASP fine volume to excess CO between fresh and aged aerosol samples. This would suggest a loss of aerosol mass from the fine mode with time, which is difficult to explain. Losses of fine volume (radii <0.35 μm) could occur by gravitational settling or by coagulation of fine particles with coarse dust particles. However, there was little evidence for internal mixing between biomass burning aerosol in electron microscope images of aerosols collected on filters [Formenti et al., 2008]. Cloud processing and wet deposition are unlikely to be responsible for particle losses since cloud was rarely observed within the aerosol layers. The apparent loss of PCASP fine volume may be partly due to sampling bias, since a threshold PCASP concentration was used to identify data corresponding to fresh smoke plumes (see section 3.1). However, this potential sampling bias cannot explain why the ratio of CNC to PCASP concentration decreases with aerosol age.
 Coagulation therefore seems the most likely explanation for the observed changes in particle number and ΔCO ratios. Condensation may also occur and may partly contribute to the increase in count mean radius. However, ratios of CO and organic mass presented by Capes et al.  suggest that condensation had a minor role in the evolution of aerosol size distribution and mass during DABEX. Furthermore, the evolution of size distribution shown in Figure 10 was simulated reasonably well by a simple model relying only on the process of coagulation (also in the work by Capes et al. ). Together, these results suggest that coagulation is more likely to be the dominant process governing the evolution of aerosol size distributions shown in this study.
6. Observed Single Scattering Albedo
 Aerosol single scattering albedos have been calculated from the nephelometer scattering coefficient and PSAP absorption coefficient for a wavelength of 0.55 μm (see sections 2, 2.1, and 2.2). Results for all runs during DABEX are shown in Figure 12, plotted against the Angstrom exponent from the nephelometer (based on the 0.45 and 0.7 μm channels). Runs with Angstrom exponents greater than 1.0 have been categorized as “dominated by aged biomass burning aerosol” (see section 3.2) and are shown by bold symbols in Figure 12. None of the runs through fresh biomass burning aerosols had Angstrom exponents above 1.0 because the proportion of scattering from dust was too high. Therefore, it has not been possible to assess the single scattering albedo of the fresh biomass burning aerosol. single scattering albedos from the runs in dominated by aged aerosol. Single scattering albedos for the aged aerosol ranged from 0.73 to 0.93. Some of this variation is likely to be related to the relative proportion of mineral dust in the aerosol sample because the single scattering albedo of the dust was observed as being much higher at around 0.98–0.99 ± 0.02. This is shown by the group of data points in Figure 12 with Angstrom exponents close to zero. The very high single scattering albedos observed in these dust runs are probably overestimated by 0.01–0.02 since losses of large particles in the PSAP instrument can lead to an underestimation of absorption, particularly in dusty conditions [Osborne et al., 2008]. Notwithstanding this uncertainty, the observations show a general trend of increasing single scattering albedo with decreasing Angstrom exponent.
 Much of the variability in single scattering albedo is however, unexplained by the Angstrom exponent, and must be due to other factors, such as black carbon content. The combustion material and combustion characteristics (intensity, temperature, efficiency, proportions of flaming and smouldering combustion) were found to influence black carbon content and single scattering albedo during SCAR-B [Reid and Hobbs, 1998]. Since there is a strong north-south gradient in vegetation density within the Sahel, it is possible that biomass burning aerosol characteristics could vary between sources areas within the region. The aircraft runs included in this analysis cover a broad geographic region, as shown in Figure 2, which may explain some of the variability. Loss of large particles through sample inlets and tubing will also influence single scattering albedos derived from the nephelometer and PSAP. This sources of error, and size-dependent corrections to the Nephelometer (see sections 2.1 and 2.2) may also lead to some of the variability in the measured single scattering albedo.
 The mean single scattering albedo of all runs dominated by aged biomass burning aerosol was 0.81 ± 0.05. The uncertainty range of ± 0.05 only includes uncertainties associated with instrumental corrections applied to the PSAP and nephelometer. Mixing of the biomass burning aerosol with dust is an additional source of uncertainty that is difficult to quantify. This uncertainty has been reduced by selecting runs with high Angstrom exponents, although the threshold of 1.0 is somewhat arbitrary and raising its value generally leads to lower single scattering albedos. The limited number of data points with Angstrom exponents over 1.0, and the spread of single scattering albedos within that group (Figure 12) are also sources of uncertainty in estimating their mean value. Aircraft single scattering albedos have been compared with independent in situ measurements from the AMMA SOP-0 ground site at Djougou, Benin [Mallet et al., 2008]. The aircraft gave values 0.02–0.06 lower than at the ground site, although the comparison is limited to only one flight. It is difficult to know whether the altitude of the aircraft (500 m above ground level) might have biased the comparison since single scattering albedos often decreased with height during DABEX [see Osborne et al., 2008]. The ground instruments are also likely to sample coarse particles more efficiently, which would affect the comparison to some degree.
 The mean single scattering albedo of 0.81 is slightly lower than reported in past measurements of aged biomass burning aerosol. For instance, a reassessment of Haywood et al. [2003a], using the identical techniques and corrections applied here, gives single scattering albedo in the range 0.83–0.92 for aged biomass burning aerosol and a mean value of 0.88. Magi et al. , gives single scattering albedo estimates of 0.81–0.9 from aircraft profiles in boundary layers over southern Africa, dominated by biomass burning aerosol. Reid et al.  report a single scattering albedo of 0.83–0.86 for regional haze dominated by aged smoke in Brazil during SCAR-B. Chand et al.  give a single scattering albedo of around 0.92 for surface observations of aerosols over the Amazon basin during the 2002 biomass burning season. See Reid et al. [2005b] for a more complete review of single scattering albedo estimates and other optical properties. Variations in single scattering albedo between DABEX and other measurement campaigns could be due to a range of factors, including mixing of biomass burning aerosol with industrial and urban emissions, mixing with mineral dust, and variability in biomass burning combustion characteristics.
7. Aerosol Optical Properties From Mie Scattering Theory
 Aerosol optical properties have been calculated for the biomass burning aerosol by Mie scattering theory for a wavelength of 0.55 μm, and are presented in Table 3. Lognormal model fits were used to represent the aerosol size distributions to make these results easier to compare with other models (see Table 1 for lognormal parameters). A density of 1.35g cm−3 is assumed, as in the work by Reid and Hobbs  and Haywood et al. [2003a]. The real part of the refractive index is assumed to be 1.54, as in the work by Haywood et al. [2003a], based on Sun photometer retrievals [Yamasoe et al., 1998]. AERONET Sun photometer retrievals from Banizoumbou during DABEX general give values around 1.53–1.59 but, in any case, aerosol optical properties are not very sensitive to the real part of the refractive index. The imaginary part of refractive index is crucial in governing shortwave absorption and is estimated to be 0.045i as this leads to a Mie-calculated single scattering albedo of 0.81, which matches the value observed for the aged biomass burning aerosol.
Table 3. Observed and Mie-Derived Optical Properties for Aged Biomass Burning Aerosol (for 0.55 μm Wavelength)a
0.81 ± 0.05
Properties include scattering coefficient from the nephelometer (σneph), and PCASP (σPCASP) in K m−1, single scattering albedo (ω), specific extinction coefficient (kext) in m2 g−1, asymmetry parameter (g), and Angstrom exponent (A) between 0.45 and 0.7 μm.
Aged (modes 1 + 2)
 The imaginary part of the refractive index 0.045i, is a factor of 2.5 higher than the value of 0.018i suggested by Haywood et al. [2003a] to characterize aged biomass burning aerosol from southern Africa. This implies a much higher black carbon content of 12.5% for DABEX, compared to 5% from SAFARI. These estimates assume mass densities of 1.35g cm−3 for the biomass burning aerosol and 1.7g m−3 for black carbon, assume a refractive index of 1.75 + 0.44i for black carbon [World Climate Program, 1986], and uses the Maxwell-Garnet rule [e.g., Chylek et al., 1988] to describe the internal mixing of the black carbon within the biomass burning aerosol. Modeling studies should be cautious in applying DABEX values for black carbon content or refractive index in global models, because of the large variability in observed single scattering albedo (see section 6). The specific extinction coefficient, asymmetry parameter and Angstrom exponent for the aged aerosol (Table 3) are similar to values from SAFARI-2000 [Haywood et al., 2003a]. This reflects similarity in observed biomass burning aerosol size distributions between southern Africa and the Sahel. Here the Angstrom exponent is derived from the extinction at 0.45 and 0.7 μm so that it is consistent with those derived from the nephelometer and shown in Figures 4 and 12. Observed (nephelometer-derived) Angstrom exponents only reach as high as 1.5. The difference between Mie-calculated and nephelometer-derived Angstrom exponent is most likely related to external mixing of the biomass aerosols with dust, greatly increasing the volume of coarse particles. In contrast, the size distributions used in the Mie calculations were cut off at 0.35 μm to exclude the influence of coarse particles on the calculated optical properties, which are assumed to be predominantly mineral dust.
 Since it was not possible to derive a single scattering albedo for the fresh biomass burning aerosol it is difficult to suggest a refractive index for the fresh aerosol. Past studies [e.g., Haywood et al., 2003a; Abel et al., 2003; Reid et al., 1998] show fresh aerosol to have higher black carbon content than aged aerosol. However, in the absence of this information it is difficult to suggest what increase in black carbon content is realistic for the DABEX aerosol. We cautiously suggest results from mode 1 as a possibility for the optical properties of fresh biomass burning aerosols during DABEX, which would assume that the black carbon content and refractive index do not vary between the fresh and aged aerosol. This would give the fresh aerosol a single scattering albedo of 0.78, which is 0.03 lower than for the aged aerosol. This tendency toward a lower single scattering albedo for the fresh aerosol is at least consistent with past studies. For example, fresh smoke during SCAR-B had single scattering albedos in the range 0.74 to 0.84 [Reid and Hobbs, 1998], whereas regional hazes from SCAR-B had single scattering albedos of 0.83–0.86. Single scattering albedos were observed to increase from 0.84 ± 0.04 to 0.90 ± 0.04 between smoke close to the source and aged haze 5 h downwind from a large fire during SAFARI-2000 [Abel et al., 2003]. Most of this change was attributed to changes in aerosol composition rather than size distribution.
Table 3 also compares the mean scattering coefficient derived from the PCASP size distribution against the scattering coefficient from the nephelometer. These are averages over the 16 runs categorized as “dominated by aged biomass burning aerosol” in Figure 4, see section 3.2. Here, the PCASP data has been used as direct input for the Mie calculations, rather than using the lognormal fits. The refractive index is assumed to be 1.54 + 0.045i for particles less than 0.35 μm, and 1.53 + 0.0004i for larger particles (0.35–1.60 μm) on the basis of Osborne et al. . The scattering coefficient derived from the PCASP (averaged over the 16 runs dominated by aged biomass burning aerosol) is only 58% of the value given by the nephelometer. This underestimation factor was fairly consistent between runs with a standard deviation of 8%. Past studies have also noted discrepancies between nephelometer and PCASP-derived scattering coefficients [Haywood et al., 2003b; Osborne et al., 2004, 2007] with PCASP-derived value between 50% and 80% of the nephelometer values. The underestimation was thought to be due to underestimation of total concentration as Haywood et al. [2003b] show an excellent agreement between the PCASP size distribution and that derived from radiometeric AERONET measurements. The underestimation was greater during profile ascents than profile descents and suggest that PCASP aerosol number concentration is affected by differences in aircraft pitch angle, which may affect the flow around the wing-mounted PCASP. The fact that the PCASP only measures particles to a nominal radius of 1.5 μm means that the PCASP measurements miss a considerable proportion of the extinction associated with the coarse mineral dust particles. Therefore, during DABEX when fine and coarse particles are mixed, determination of the scattering or extinction from PCASP is not considered reliable. However, we have reasonable confidence in the shape of the PCASP size distribution from comparisons of the PCASP against AERONET retrievals, and other in situ measurements in Haywood et al. [2003b], Osborne et al. , and Osborne et al. , plus the similarity between DABEX and SAFARI size distributions shown in Figures 9 and 11.
 The physical and optical properties of biomass burning aerosols have been characterized using data from the FAAM aircraft over West Africa during DABEX (January–February 2006). Aged biomass burning aerosols were observed on every flight and typically found at altitudes of 1.5 to 4 km. Dust was typically observed below 1.5 km but also at higher altitudes, mixed with biomass burning aerosol. Fresh biomass burning aerosols were observed, south of Niamey (10°N), by flying at approximately 300 m over regions with numerous agricultural fires. Smoke plumes are identified in this data by sharp peaks in aerosol concentration.
 Biomass burning aerosols were mainly composed of fine particles with radii smaller than 0.35 μm. The fresh aerosol had high concentrations of fine particles with radii in the range 0.05–0.1 μm, whereas the aged aerosol had fewer of these fine particles and more particles with radii of 0.1–0.2 μm. The aerosol size distributions are represented well by a two-mode lognormal distribution. The fresh aerosol is represented well by a single lognormal distribution with geometric mean radius of 0.08 μm and standard deviation of 1.4. The aged aerosol is represented best by two lognormal modes, the first mode having the same parameters as the fresh model fit, the second mode having a geometric mean radius of 0.16 and a standard deviation of 1.25. Results show a substantial drop in the ratio of aerosol number concentration to CO, indicating a substantial rate of coagulation. This process is likely to have played a major role in the evolution of the size distribution, although condensation may also have contributed. The observed aerosol size distributions are broadly similar to observations made during past field experiments over southern Africa and Brazil. This is encouraging for modeling applications.
 The aerosol single scattering albedo was extremely variable during DABEX; values ranged from 0.73–0.93 for layers dominated by biomass burning aerosol (all values quoted for a wavelength of 0.55μm). The single scattering albedo generally increased with decreasing Angstrom exponent as low Angstrom exponents were indicative of a high dust proportion, and the dust was less absorbing than the biomass burning aerosol. However, we attribute much of the variability of single scattering albedo to other factors, possibly source region, combustion type, and the possible influence of internal mixing with dust. The mean single scattering albedo for the aged biomass burning aerosol was estimated as 0.81 ± with an instrumental uncertainty of 0.05. This estimate will inevitably be influenced to some degree by mixing with mineral dust, although this influenced was minimize as far as possible by averaging over data where the Angstrom exponent was greater than 1. The single scattering albedo for the fresh aerosol was difficult to determine because the dust loading was too high during the fresh smoke flights and would have heavily biased the optical properties.
 A refractive index of 1.54 + 0.045i was required to gain agreement between the observed single scattering albedo and Mie calculations based on the lognormal size distribution for the aged aerosol. This refractive index implies a black carbon mass content of 12.5% for the aged aerosol. This level of black carbon and shortwave absorption is much higher than observed for aged biomass burning in past field experiments over southern Africa and South America. This difference may be important for regional climate since a higher black carbon content (or lower single scattering albedo) will lead to a more positive direct radiative forcing. In the Mie calculations, the aged biomass burning aerosol had a specific extinction coefficient of 5.8 m−2 g−1, an asymmetry parameter of 0.63, and an Angstrom exponent of 1.73. These measurements will be useful for both modeling efforts and for the validation of aerosol parameters assumed in remote sensing retrievals.
 The FAAM aircraft is jointly funded by the Met Office and the Natural Environment Research Council.