Size distribution, shape, and composition of mineral dust aerosols collected during the African Monsoon Multidisciplinary Analysis Special Observation Period 0: Dust and Biomass-Burning Experiment field campaign in Niger, January 2006



[1] Dust samples were collected onboard the UK community BAe-146 research aircraft of the Facility for Airborne Atmospheric Measurements (FAAM) operated over Niger during the winter Special Observation Period of the African Monsoon Multidisciplinary Analysis project (AMMA SOP0/DABEX). Particle size, morphology, and composition were assessed using single-particle analysis by analytical scanning and transmission electron microscopy. The aerosol was found to be composed of externally mixed mineral dust and biomass burning particles. Mineral dust consists mainly of aluminosilicates in the form of illite and kaolinite and quartz, accounting for up to 80% of the aerosol number. Fe-rich particles (iron oxides) represented 4% of the particle number in the submicron fraction. Diatoms were found on all the samples, suggesting that emissions from the Bodélé depression were also contributing to the aerosol load. Satellite images confirm that the Bodélé source was active during the period of investigation. Biomass burning aerosols accounted for about 15% of the particle number of 0.1–0.6 μm diameter and were composed almost exclusively of particles containing potassium and sulfur. Soot particles were very rare. The aspect ratio AR is a measure of particle elongation. The upper limit of the AR value distribution is 5 and the median is 1.7, which suggests that mineral dust particles could be described as ellipsoids whose major axis never exceeds 1.9 × Dp (the spherical geometric diameter). This is consistent with other published values for mineral dust, including the recent Aerosol Robotic Network retrieval results of Dubovik et al. (2006).

1. Introduction

[2] Mineral dust is one of the most abundant aerosol species in the atmosphere in terms of emitted mass [Forster et al., 2007]. One of the major climatic effects of mineral dust aerosols is due to their efficiency in scattering and absorbing solar and terrestrial radiation. However, the quantification of the effect of mineral dust on the Earth-atmosphere radiative budget is prone to uncertainties due to the underdetermination of its optical properties, which, in turn, is caused by a poor knowledge of its intrinsic properties, such as the elemental composition, the number size distribution, and shape.

[3] This paper discusses the elemental composition, size distribution, and shape of individual mineral dust particles collected in Niger on board the Facility for Airborne Atmospheric Measurements (FAAM) BAe-146 research aircraft. The aircraft was operated from Niamey (Niger) between 10 January and 3 February 2006 in the framework of the first Special Observation Period (SOP0) of the African Monsoon Multidisciplinary Analyses (AMMA) project [Redelsperger et al., 2006]. This experiment aimed to assess the direct radiative impact of mineral dust at the regional scale of western Africa. Africa is the largest world source of mineral dust: total annual Saharan emission of 670 Mt have been estimated by Laurent [2005]. In wintertime, biomass burning emissions are also intense, and mixing between the two components might alter their optical properties.

[4] Single-particle analysis by electron microscopy has been widely used to assess properties of aerosols, especially mineral dust [Falkovich et al., 2001; Okada et al., 2001; E. A. Reid et al., 2003]. Electron microscopy is the preferential tool to study the shape of aerosols; furthermore, it allows partitioning the elemental composition by number and not by mass as it is done by bulk analysis techniques. This is needed to calculate optical properties.

[5] This paper completes the body of observations on dust composition and optical properties presented by Formenti et al. [2008] and Osborne et al. [2008].

2. Methodology

2.1. Sample Collection and Preparation

[6] The aerosol sampling system used on board the BAe146 is the same that has been previously used on board the UK Met Office C-130 to characterize mineral dust (e.g., the Saharan Dust Experiment (SHADE)) [Tanré et al., 2003]. It is described in detail by Andreae et al. [2000]. The aerosol inlet consists of a thin-walled inlet nozzle with a curved leading edge; the design was based on criteria for aircraft engine intakes at low Mach numbers [Andreae et al., 1988]. This design reduces distortion of the pressure field at the nozzle tip and the resulting problems associated with flow separation and turbulence. The sampling system operated at flow rates that averaged 120 L min−1 (at ambient pressure and temperature); the flow was adjusted to maintain slightly subisokinetic sampling conditions. The aerosol intake system was designed so that rain and large cloud water droplets would be removed from the sampled air stream by inertial separation.

[7] Aerosol particles were sampled by filtration onto two stacked-filter units (SFUs) mounted in parallel. Each SFU can hold a maximum of three filters on sequential 47-mm diameter polyethylene supports, but only one stage was used during AMMA. Samples were collected only during horizontal flight legs lasting not less than 20–30 min in order to guarantee sufficient loading of the filter samples. Each SFU consisted of a Nuclepore filter of 90-mm diameter and nominal pore size 0.4 μm. One SFU was used for measuring water-soluble ions and major, minor, and trace elements. The second SFU was used for measuring carbonaceous aerosols on Whatman QMA quartz filters. Blank samples were collected on every flight by placing filters in the sampling line as if they were actual samples and exposing them to the air stream for a few seconds.

[8] Immediately after each flight, both loaded and blank filters were stored in Petri dishes. Back in the laboratory, samples were cut in subportions of various surface areas (from one eighth up to one half of the filter) which were then analyzed with various techniques. Further details are given by Formenti et al. [2008]. One eighth of the filter was dedicated to electron microscopy analysis. In order to make the filter substrate (polycarbonate) conductive, samples were coated either with platinum or with carbon.

2.2. Sample Analysis

[9] Electron microscopy analysis is used to investigate the size distribution, the shape, and the composition of individual aerosol particles. As recommended by E. A. Reid et al. [2003], we use a combination of scanning and transmission electron microscopes, both equipped with an energy dispersive X-ray detection system. This allows investigating the 2-D structure of particles both in the fine and in the coarse size fractions, with some information on depth and 3-D structure for larger particles.

[10] Analytical scanning electron microscopy (SEM) is performed with an instrument type JEOL 6301F equipped with an X-ray energy-dispersive spectrometer (Oxford Link Pentafet Detector and Link ISIS analyzer, Oxford Instruments, UK). SEM allows obtaining “bulk” images of aerosol particles, that is, 2-D images with some information regarding their volume. This is particularly useful in the investigation of mineral dust which consists generally of complex aggregates in the coarse fraction. SEM has a large dynamic range of magnification as it allows to image particle of diameter from fractions to tenths of micron.

[11] Analytical transmission electron microscopy (TEM) is performed with an instrument type JEOL 100CXII equipped with a X-ray energy-dispersive spectrometer (PGT Prism 2000 Si(Li) detector and Avalon analyzer, Princeton Gamma-Tech, USA). TEM provides a better resolution for particles smaller than 0.5 micron in diameter. However, only 2-D images of particles are provided without any further information on their bulk structure.

[12] Regarding size and shape, we acquire images at various magnifications in order to investigate the largest size spectrum as possible. These are analyzed using the HISTOLAB counting program (Microvision Instruments, France). On the basis of the light contrast between the particles and the background filter, the particle projected area Aproj is measured as the number of contiguous pixels whose brightness is higher than a predetermined threshold value corresponding to the filter background. The particle geometric diameter Dp is calculated from the projected area Aproj assumed as the equivalent area of a spherical particle

equation image

[13] Images are inspected manually in order to establish the best settings for automatic particle detection (threshold of brightness and contrast). Manual adjustment is necessary in order to take into account particles whose contrast is too low to be detected by the algorithm (negative artifact) or to discard those which are only partially included in the field images (positive artifact).

[14] This semiautomatic analysis provides with the total number of particles per image (Ni,TOT) as well as their number size distribution dNi(Dp). The particle number size distribution in the atmosphere dN(Dp) can then be calculated as

equation image

where Sfilter is the surface of the filter STOT is the total analyzed surface Vair is the air volume sampled through the filter. In the following, the number size distribution are presented as dN(Dp)/dlogDp.

[15] Aerosols, and in particular mineral dust, are not spherical particles but are characterized by a noneven shape distribution as a function of size. In order to characterize the particle morphology we use two parameters: the aspect ratio AR and the shape factor Sh [E. A. Reid et al., 2003]. The aspect ratio AR is a measure of the sphericity of a particle and it is defined by the formula

equation image

where Lproj is the major projected dimension of the particle. The aspect ratio AR is equal to one for a spherical particle and greater than 1 for elongated particles such as ellipsoids.

[16] The shape factor Sh provides information about the particle roughness and discontinuities and it is defined as

equation image

where P is the particle perimeter. Like the aspect ratio, the shape factor Sh is equal to one for a spherical particle and larger than one for aggregates and irregular particles.

[17] Finally, the elemental composition of individual particles is obtained by energy dispersive X-ray fluorescence spectrometry. TEM is used to measure the elemental composition of particles of geometric diameter <1 μm, which are chosen randomly on the whole of the filter. SEM is used to measure the elemental composition of particles of geometric diameter >1 μm. This is done by selecting two images at a convenient magnification and by analyzing all particles found here within. The live time analysis is 80 s of TEM and 50 s for SEM. Because the analysis is done by manual selection of the particles, only 300 particles have been analyzed by TEM and 490 particles by SEM.

2.3. Ancillary Data

[18] Additional information regarding the particle morphology has been obtained by the Atomic Force Microscopy (AFM) using the instrument available at the Max Plank Institute for Chemistry in Mainz, Germany (CP-Research, ThermoMicroscopes, now Digital Instruments, Veeco). The AFM can provide information regarding the three-dimensional particle structure which is not displayed by SEM and TEM. However, only a very limited number of particles could be investigated as this type of analysis is rather time consuming. Furthermore, owing to the technique limitation [Gwaze et al., 2007], only particles whose heights are lower than 1.5 μm could be investigated.

[19] Aerosol particle size distributions were also measured with two passive cavity aerosol spectrometer probes (PCASP-100X, PMS Inc., Boulder, Colorado), one wing-mounted and one sampling from a Counterflow Virtual Impactor (CVI) inlet (operating in direct-sample aerosol mode) located inside the cabin (CVI-PCASP in the following). The wing-mounted PCASP is the standard single particle PMS counter classifying particles into 15 size channels in the diameter range between 0.1 and 3 μm (nominal). The measuring range of the CVI-PCASP is 0.1–10 μm diameter. Further details are given by Osborne et al. [2008].

[20] Dispersion modeling was undertaken using the Met Office Numerical Atmospheric-Dispersion Modeling Environment (NAME). This is a Lagrangian particle model [Ryall and Maryon, 1998] in which emissions from pollutant sources are represented by parcels released into a model atmosphere driven by the meteorological fields from the Met Office's numerical weather prediction model, the Unified Model [Cullen, 1993].

[21] Two types of model output are presented in Figure 1. Back trajectories (Figure 1, left) indicate the origin of the air arriving at the plane while the samples were taken. The output shown here represents the movement (both in the horizontal and vertical) of 99 air parcels. The second model output presented here (Figure 1, right) was generated using 100,000 air parcels to represent the origin of the air arriving at the plane. Output has been restricted to those air parcels present within 500 m of the surface during transit. As such this represents the source areas that are likely to have contributed to the dust/biomass burning aerosol observed by the plane.

Figure 1.

Airflow information by backward trajectories and dispersion modeling for samples B160N3, B161N3, B161N5, and B165N7. (left) Three-day back-trajectories have been calculated using the NAME model by releasing 99 air parcels over the time period between the start and end position (latitude, longitude, and elevation) of each of the sample SLR runs. (right) Backward dispersion simulations have been performed using the NAME model using 100,000 air parcels. Plots show the contribution of surface air masses (integrated below 500 m) to the aircraft position (red point in the maps). The Ozone Monitoring Instrument (OMI) on Aura daily images of the UV aerosol index for the day prior sampling and day of sampling are reported in order to identify areas of elevated dust load which might correspond to dust sources having contributed to the collected aerosol samples. Instantaneous dust product maps from the SEVIRI satellite are reported to illustrate that regional extent of the aerosol load at the time of sampling.

Figure 1.


Figure 1.


Figure 1.


3. Aerosol Samples and Air Mass Origin

[22] Fourteen research flights were performed by the BAe-146 during AMMA SOP0. The operating regions and flight tracks are described by Haywood et al. [2008]. In this section we present the summary of sampling conditions (sample ID, dates, daytime, duration of sampling, geographic position (latitude, longitude, and elevation), and estimated dust concentration, see Table 1) for the four samples discussed in this paper. Further information regarding the aerosol vertical profile is provided by Osborne et al. [2008]. Indication of the air mass origin and the likely source regions are presented in Figure 1.

Table 1. Sample ID, Dates, Sampling Duration, Minimum and Maximum Geographic Position, and Estimated Dust Concentration for the Samples Under Discussion in This Paper
Sample IDDateSampling Duration (min)Geographic PositionEstimated Dust Concentration (μg m−3)
Latitude (degN)Longitude (degE)Elevation (m agl)
B160N321 Jan 200653:4916–18.54.9–6.7600–1500360
B161N323 Jan 200610:0516.4–17.26.1–7.11000590
B161N523 Jan 200615:3017.4–16.87.5–6.610001234
B165N730 Jan 200614:2313.62.8–1.9150464

[23] Sample B160N3 was collected on 21 January 2006 during a straight and leveled run (SLR) lasting about 1 h in a dust layer over the northeast of Niger. During this long run, the aircraft altitude was adjusted twice between 600 and 1500 m above ground level (agl) in order to maximize the filter loading. Dust concentration was not measured directly, but as a first approximation, it can be estimated from the elemental Al concentration measured by particle induced X-ray emission (PIXE) analysis of one of the filter subportions [Formenti et al., 2008], knowing that Al is about 8%. When doing so, the dust concentration was 360 μg m−3. Back trajectories indicate an easterly flow. More specific modeling indicated that the surface sources of aerosol mainly come from Niger, Chad (including the Bodélé Depression), Sudan, and Egypt. In addition there are potential surface sources south of Niger which may include biomass burning and anthropogenic aerosols.

[24] Samples B161N3 and B161N5 were collected on 23 January 2006 during reciprocal runs at constant height (∼1000 m agl), in correspondence with the maximum dust concentration (590 and 1230 μg m−3, respectively). Back trajectories show that air masses came mainly from the east with a smaller contribution from the north and west. Considering the sources near the surface, however, indicates that the dominant source region is Niger itself extending north to Algeria. A contribution from the east (Chad, Sudan, Egypt) remains but is less dominant (Figures 1b and 1c).

[25] B165N7 was collected on 30 January 2006 in the vicinity of Niamey at approximately 150 m agl. Unlike the others samples, in this case elevated dust concentration was encountered from the surface to 1000 m agl (460 μg m−3). Back trajectories indicate that the origin of the air mass was split between a contribution from the east and a more local diffuse contribution from the west. Plots considering only the low level sources confirm the local source to the west, mainly confined to Mali (Figure 1d).

4. Results

4.1. Sample Homogeneity

[26] In order to make sure that results obtained on a filter segment are representative of the aerosol in the atmosphere, we try and quantify the homogeneity of the aerosol deposit on the filters. To do so, we selected a sample collected on a dust layer (B163N5, dust concentration 410 μg m−3). We analyzed images acquired on various portions of the filter along a radial from the center to the edge. As suggested by visual inspection of the acquired images (not shown), the particle distribution is rather homogeneous. The variability in the total particle number per image was found to be within 15%. This result was then confirmed by repeating the analysis on each of the analyzed samples.

4.2. Single Particle Elemental Composition

[27] The single particle elemental composition by energy dispersive X-ray spectrometry has been investigated separately in the fine fraction (particles of diameter smaller than 1 μm) and in the coarse size fraction (particles of diameter larger than 1 μm). Overall, the elemental composition of the various samples is rather homogeneous. Only two aerosol types are encountered, mineral dust and aerosols of anthropogenic origin, likely biomass burning. As shown in Figure 2, these two components are found to be externally mixed.

Figure 2.

TEM image of B160N3 sample showing externally mixed mineral dust and biomass burning K–S particles.

[28] Particles could be classified into eight categories described below. Examples of images and associated X-ray spectra are shown in Figure 3.

Figure 3.

SEM and TEM images and associated X-ray spectra of individual particles encountered on the samples: (a) illite, (b) kaolinite, (c) quartz, (d) diatoms, (e) calcium-rich, (f) iron-rich, (g) K-S and soot particles. See text for explanation. The platinum X-ray peak in the SEM spectra is due to the sample preparation, whereas the copper line in the TEM spectra is due to the sample holder. The lines of carbon and oxygen are visible but mainly due to the filter substrate.

Figure 3.


[29] 1. Aluminosilicates characterized by the dominance of Al and Si were found both in the fine and in the coarse size fractions, although rarely below 0.4 μm diameter. Their identification is often complicated but can be made on the bases of both chemical and morphological information. They were mostly illite (Si-to-Al signal ratio of 2), kaolinite (Si-to-Al signal ratio of 1), or feldspath (Si-to-Al signal ratio of 3). Trace elements were Fe and more seldom Ca and K. Examples of images are shown in Figures 3a and 3b. Phyllosillicate minerals such as montmorillonite and chlorite were observed.

[30] 2. The silicon-rich group is characterized by particles whose X-ray signal is dominated by Si (relative X-ray intensity > 75%). Just like aluminosilicates, silicon-rich particles were found both in the fine and in the coarse fractions, and their diameter was rarely lower than 0.4 μm. Silicon-rich particles were mostly crystalline quartz (SiO2, Figure 3c) but also amorphous fossil diatoms. These are microorganisms which are generally found in dry lake bed such as of the Bodélé depression [Todd et al., 2007]. Diatoms were observed generally as fragments, suggesting that they already spent a considerable time in the atmosphere at the time of collection. However, well preserved specimens were also observed, as shown in Figure 3d.

[31] 3. Calcium-rich particles (Figure 3e) are characterized by a relative intensity of the Ca X ray > 60%. They were generally simple calcium carbonate in the form of CaCO3, whereas no dolomite particles were observed (CaMg(CO3)2).

[32] 4. The Ca-S group is composed of particles for which the ratios of calcium and sulfur are equivalent. These are found more often in the coarse fraction. Their origin can be primary as gypsum (Ca-SO4·· 2(H2O)) or secondary as Ca-SO4·formed in the atmosphere in the reaction between CaCO3 and SO2 or H2SO4.

[33] 5. The iron-rich group (Figure 3f) is composed by particles for which the X-ray relative intensity of Fe dominated the signal. These are mainly iron oxide or hydroxide particles, i.e., hematite (Fe2O3) or goethite (FeO · OH) and are found more often in the fine fraction.

[34] 6. Titanium-rich particles in the form of TiO2 (rutile). These were rare but were found both in the fine and the coarse fractions.

[35] 7. Particles enriched in K and S (Figure 3g). These are almost spherical particles looking like empty shells which were found exclusively in the fine fraction (diameter 0.1–0.6 μm). Traces of Ca were sometimes detected. As discussed in the following (section 5) these particles are attributed to biomass burning.

[36] 8. Soot particles made of carbon could not be identified based on their chemical composition but were recognized based on their morphology. The dominant shape was that of elongated chains (Figure 3g). They were found exclusively in the fine fraction and in very low amounts.

[37] Chlorine-containing particles such as halite (NaCl) or KCl salts from biomass burning were not detected in the samples. This is consistent with findings on the bulk elemental composition by particle induced X-ray emission (PIXE) analysis of the same samples [Formenti et al., 2008].

[38] The mean elemental composition by number in the submicron fraction is shown in Figure 4a. Mineral dust was composed by aluminosilicates (61% (±10) by number), whose 57% (±8) were illite and kaolinite and 4% (±3) were other phyllosillicate clays. Si-rich particles accounted for 17% (±12), the variability being due to the elevated Si-rich particles observed on sample B160N3 (32%) and the strangely low percent observed on sample B161N3 (4%). Iron-rich particle accounted for 4% (±1), calcium-rich for 2% (±2), titanium-rich for 1% (±1), and Ca-S particles for 1% (±2).

Figure 4.

Mean elemental composition partitioned by number (percent) obtained for the AMMA SOP0 samples B160N3, B161N3, B161N5, and B165N7: (top) fine fraction (submicron particles); (bottom) coarse fraction (supermicron particles).

[39] Regarding combustion aerosols, K-S particles represented 15% (±3) of the total particle number. The number of soot particles was very low, less than 1% of the total number of particles. The upper limit was obtained for sample B165N7, which was collected in the vicinity of Niamey, where anthropogenic activities might be more important [Osborne et al., 2008].

[40] Figure 4b shows the mean elemental composition by number in the supermicron fraction. Here only mineral dust was found. The composition of the coarse mode dust was remarkably uniform from sample to sample. When averaged over the different samples, aluminosilicates accounted for 78% (±1) by number, of which 74% where illite and kaolinite and 4% were other phyllosillicate clays. The Si-rich group accounted for 18% (±3), Ca-rich particles for 3% (±1). Ca-S particles were found on two of the samples where they accounted for 1% of the particle number.

4.3. Particle Number Size Distribution

[41] Overall, about 31,000 particles were analyzed. Particles were classified over 19 size classes ranging from 0.01 to 7 μm for particles observed with the TEM and 17 size classes ranging from 0.25 to 10 μm for particles observed with the SEM. The TEM and SEM number size distributions are shown in Figure 5 where they are compared with those measured in situ by the two PCASP counters, the extended range operating from inside the cabin (CVI-PCASP) and that wing-mounted. In their paper, Osborne et al. [2008] found that the CVI-PCASP consistently overcounts relative to the wing PCASP, likely due to a problem with the sample flow on the CVI PCASP. In order to extend the size range of measurements, these authors rescaled the CVI-PCASP data to match the PCASP ones in the region of overlap (<3 μm diameter), then they projected the CVI-PCASP large particle counts in the range 3–10 μm onto the PCASP distribution, i.e., they extended the wing PCASP data. However, for the purpose of the present paper, we decided to leave the CVI-PCASP data uncorrected and regard the differences between the CVI-PCASP and the wing PCASP counts as the range of uncertainty in the OPC particle counting. All distributions are represented as dN/dlogDp.

Figure 5.

Comparison of number size distributions dN/dlogDp obtained by electron microscopy (TEM + SEM) and by optical counting (CVI PCASP and PCASP). The best (sample B160N3) and the worst (sample B165N7) case studies are shown. On the x axis the diameter is geometrical in the case of the TEM + SEM results and optical for measurements by the optical counters.

[42] The agreement is good for particles smaller than 0.5 μm, which are measured by the CVI-PCASP, the PCASP, and by the TEM. For particles larger than 0.5 μm, the agreement for the size distributions obtained by TEM and SEM is very good. This confirms the reliability of the methodology as the two techniques were applied on different filter portions and use different sample preparation protocols. The agreement with the optical counters is less satisfactory, although generally achieved within the experimental standard deviation. The worst agreement is obtained for sample B165N7 (Figure 5b), which is the sample more likely influenced by anthropogenic particles. Reasons for enhanced disagreement are not clear.

[43] Finally, the number size distributions obtained by combination of TEM between 0.05 and 0.5 μm and SEM between 0.5 and 10 μm are shown in Figure 6. Distributions are rather similar, with the exception again of sample B165N7, which presented a higher number of particles below 0.5 μm. Note that this also suggests that the coarse particle number measured by SEM in sample B165N7 is not overestimated relative to the other samples, as it could have been suspected from Figure 5b.

Figure 6.

Number size distributions obtained by combination of TEM between 0.05 and 0.5 μm and SEM between 0.5 and 10 μm.

[44] In addition, the fit with a three-mode lognormal distribution model shows that sample B160N3 has a mode in the coarse fraction at 4.4 μm diameter, whereas the coarse mode distribution peaks at 1.5–2.0 μm for the other samples. Parameters of the log-normal fit distribution are shown in Table 2.

Table 2. Three-Mode Lognormal Parameters for Number Size Distributions Obtained by Combination of TEM Between 0.05 and 0.5 μm and SEM Between 0.5 and 10 μm
NTOT (cm−3)Dp (μm)σNTOT (cm−3)Dp (μm)σNTOT (cm−3)Dp (μm)σ

4.4. Particle Shape Distribution

[45] Cumulative probability plots of aspect ratio AR as a function of particle diameter are presented in Figure 7a. For sake of comparison, particle diameters have been averaged over five size classes as done by E. A. Reid et al. [2003].

Figure 7.

Cumulative probability plots of (top) particle aspect ratio AR and (bottom) particle shape factor Sh.

[46] For all the samples, the AR is almost independent of particle size. Values of AR never exceed 5, that is, particles of diameter smaller than 10 μm can be described as ellipsoids with a major axis lower than 2.2 times the geometric diameter. The median AR was 1.7.

[47] Cumulative probability plots of shape factor Sh as a function of particle diameter are represented in Figure 7b. In contrast to the aspect ratio, the shape factor seems to be influenced by particle size. The median Sh decreases with particles size, indicating that smaller particles are more regular than larger ones. This is consistent with visual inspection of the particle shape indicating that irregular aggregates are encountered more often in the coarse than in the fine fraction.

[48] Although sparse, AFM measurements performed on three of the Niger samples support these observations (Figure 8). Smaller particles are rounder and smoother, whereas larger ones (diameter greater than 1 micron) look rough with many discontinuities. The AFM images also indicate that supermicron particles are rather elongated and flat; that is, their height is about one third of their major axis. This is consistent with the fact that aluminosilicates, which are in fact rather flat particles, account for more than 50% of the total particle number.

Figure 8.

AFM micrography showing (top) the particle section analyzed and (bottom) the height profile of (left) a submicron particle as well as (right) a coarse particle.

5. Discussion

[49] Data regarding the elemental composition, size distribution and shape of individual mineral dust particles collected over Niger have been presented. On this basis, the following conclusions can be made:

5.1. Aerosol Composition

[50] In the fine fraction, the aerosol was found to be composed of mineral dust and combustion aerosols in external mixing, and in rather constant proportions (85% (±3) and 15% (±3) for mineral dust and biomass burning aerosol, respectively). In the coarse fraction, dust accounts for all particles sampled.

5.2. Mineral Dust Composition and Origin

[51] The average composition (by number) of mineral dust is overall rather similar from sample to sample, particularly in the coarse fraction, and consistent with previous findings for mineral dust [Falkovich et al., 2001; E. A. Reid et al., 2003; Sobanska et al., 2003; Kandler et al., 2007]. This suggests that the source region is constant across the different flights, or if there were different source regions, the compositional fingerprinting of these different sources is masked within the experimental uncertainties. The NAME dispersion model indicates that transport of surface air masses to the sampling site was equivalent for the various samples (Figure 1). Hot spots in aerosol concentrations were investigated by satellite imagery. The SEVIRI satellite dust product images for the day of sampling (Figure 1) indicate that the dust load over the area of sampling was rather diffuse. The examination of the complete 15-min time series (available on indicates that active sources were the Bodélé region, with a peak in intensity on 21 January 2006, but also on 30 January 2006, and sources in the north of Niger (around the Aïr mountain chain) on 23 January. This is confirmed by the OMI UV absorbing index aerosol images for the day prior sampling and day of sampling (Figure 1). Indications of detection of dust from Bodélé were found in the enrichment in silicon rich-particles (quartz and diatoms), particularly on sample B160N3. Diatoms detected on sample B160N3 were much better preserved than on the other samples (Figure 3c), and they can unambiguously be related to the planktonic centric diatom Aulacoseira and Stephanodiscus species which have being shown as characteristic of Bodélé mineral dust [Todd et al., 2007; Moreno et al., 2006]. In addition, the number size distribution coarse mode peaked at a much larger value than for the other samples. As a matter of fact, back trajectories indicate that the air masses were transported directly from Bodélé to the sampling point. Finally, iron-rich particles were identified in the submicron fraction. These are iron oxides, either hematite and/or goethite, which are found either isolated or, more frequently, in aggregation with clay minerals [Lafon et al., 2006; Kandler et al., 2007]. Their percent contribution was lower in sample B160N3, which is consistent with the expected low iron content in Bodélé dust. A word of caution is necessary due to the difficulty of observing iron oxides by electron microscopy.

5.3. Biomass Burning Composition and Origin

[52] Biomass burning aerosols were observed in the fine fraction only and were almost exclusively composed by particles enriched in potassium and sulfur. These are combustion-produced particles which have been identified in previous studies in smoldering African smoke [Gaudichet et al., 1995; Liu et al., 2000]. These authors have shown that whereas K is mainly associated with chlorine in flaming smoke, it occurred predominantly as K-S particles in smoldering and ambient haze due to release of Cl in the KCl particles by acidification reactions with SO2, also generally emitted by fires. This is consistent with the fact that the biomass burning encountered over Niger during the research flights discussed in this paper is regional haze which had recirculated and thus aged. However, it is worth mentioning that the conversion from KCl to K-S already occurs at a very early stage within rather short distance from the fires [Gaudichet et al., 1995]. By applying the approach of Falkovich et al. [2001] consisting of looking at the slope of the scatterplot between the estimated elemental mass Me and the particle diameter Dp, indicates that sulfur is distributed on the surface rather than on the particle volume (MeDp1.7), which is consistent with its hypothetic origin in condensation. By contrast, slopes of mineral dust constituents such as Mg, Al, Si, Fe, K, and Ca range from 2.4 to 3.2, rather in agreement with the expected bulk distribution. Sulfur associated with mineral dust particles was observed very rarely in the fine fraction and practically never in the coarse fraction. This supports the conclusion that mixing between biomass burning and mineral dust is external. The occurrence of soot chains, which could be identified based on their characteristic morphology, was extremely occasional. Soot accounted at the most for 1% of the total particle number, this upper limit being obtained for sample B165N7 collected in the vicinity Niamey, likely a point source of pollution (car traffic, domestic burning). Back trajectories indicate that this southernmost sample is the one more influenced by air masses originating over source regions of fires, that is, Benin and Nigeria. These facts indicate that the dust plumes investigated in this paper were embedded into a sort of regional biomass burning aged haze, probably air masses which had recirculated from areas affected by active fires (Benin and Nigeria). This pollution is also evident in the enhancement of CO mixing ratios, as observed by Osborne et al. [2008]. The possibility that airflow from North Africa could contribute with polluted air masses from Egypt or even from Europe is suggested by the dispersion model results, but it is considered speculative at this stage.

5.4. Particle Shape

[53] The aspect ratio AR and shape factor Sh of about 31,000 particles were determined. The main conclusion is that the aspect ratio for particles of diameter in the range 0.1–10 μm is practically independent of size. The upper limit of the AR value distribution is 5 and the median is 1.7. This is consistent with the median value of 1.64 published by Kandler et al. [2007] but lower than that of 1.9 provided by E. A. Reid et al. [2003] for African mineral dust collected over the Caribbean after transport over the North Atlantic Ocean. Differences might not be significant due to the inherent ambiguity in retrieving diameter and shape for irregular particles from 2-D images. This is consistent with the newest retrieval results of the column volume distribution based on Aerosol Robotic Network Sun-photometer measurements indicating that nonspherical particles with aspect ratios ∼1.5 and higher dominate in desert dust plumes [Dubovik et al., 2006]. In conclusion, mineral dust particles could be described as ellipsoids whose major axis never exceeds 2.2 × Dp (the spherical geometric diameter). However, the shape factor is size-dependant. Smallest particles have more regular contours, whereas largest particles are often irregular aggregates, with rougher contours. It is currently difficult to take this parameter in account in the optical properties calculation.

5.5. Particle Number

[54] Beside losses in inlets and tubing, many artifacts affect the number size distribution by both electron microscopy and optical counting [J. S. Reid et al., 2003]. The number size distributions obtained by the two techniques could be compared in terms of modal diameter and concentrations. First al all, it is important to highlight that the filter sampling proves itself efficient in collecting particles up to tenths of microns in diameter. Differences between the filter and the OPC-based number size distributions were observed in the particle range above 0.5 μm, where the number of particles obtained by electron microscopy largely exceeds that obtained by optical counting. This might be at least partly related to an artificial enhancement of the coarse mode particle concentrations due to the fact that filter sampling is subisokinetic. However, the consequences of this discrepancy in the calculation of optical properties are minimized by the fact that the particle composition remains practically constant within the submicron and uppermicron size domains, respectively. As the shape of the size distributions is rather equivalent (that is, the modal diameters are practically the same), we can associate the size-resolved elemental composition and shape parameters obtained by optical microscopy to the number size distribution by more widespread and easy-to-use instruments such as the PCASP.

6. Concluding Remarks

[55] This paper present the first available data set of size-resolved data on elemental composition, number concentration, and shape parameters of individual mineral dust particles issued from some of the most active source regions in Africa, that is, of the globe.

[56] Source regions, identified with the help of back trajectories, dispersion calculations, and satellite images, include the Bodélé depression and the deserts of the north of Niger. Some unique features in the composition and size of collected particles could be related to a specific source region (e.g., the presence of diatomite in the Bodélé aerosol).

[57] Nonetheless, our results suggest that dust has remarkably uniform properties, particularly in terms of elemental composition and shape. This might be due to the fact that as suggested by the SEVIRI satellite images, dust emitted by localized emission hot spots converges into a widespread and diffuse load once in the atmosphere. If confirmed, this is indeed a notable simplification in the representation of transported mineral dust in regional and global models and satellite algorithms.

[58] Thanks to the AMMA program, we are able to pursue the investigation of regional variability of mineral dust properties. The analysis conducted in this paper will be repeated on samples collected during other field phases of the project, when we were able to sample mineral dust from different sources such as Mauritania, Mali, and southern Niger.

[59] Finally, this data set, providing the size-resolved elemental composition and shape of mineral dust particles by number and not by mass as it obtained by bulk analysis, is suited to estimate their optical properties both in the visible and in the infrared.


[60] On the basis of a French initiative, AMMA was built by an international scientific group and is currently funded by a large number of agencies, especially from France, UK, United States, and Africa. It has been the beneficiary of a major financial contribution from the European Community's Sixth Framework Research Programme. FAAM is jointly funded by the Natural Environment Research Council and the Met Office. The financial support of the API-AMMA French national program is acknowledged. Finally, the authors also wish to thank the BAe-146 air and ground crews, as well as the FAAM and Met Office observers, and the AMMA SOP0/DABEX PI J. Haywood (Met Office).