ARTICLE IN PRESS

1352-2310/$ - se

doi:10.1016/j.at

�CorrespondE-mail addr

Atmospheric Environment 41 (2007) 8058–8074

www.elsevier.com/locate/atmosenv

Chemical composition and complex refractive index ofSaharan Mineral Dust at Izana, Tenerife (Spain) derived

by electron microscopy

Konrad Kandlera,�, Nathalie Benkera, Ulrich Bundkeb, Emilio Cuevasc,Martin Eberta, Peter Knippertzd, Sergio Rodrıguezc,e,

Lothar Schutzd, Stephan Weinbrucha

aInstitute of Applied Geosciences, Darmstadt University of Technology, Schnittspahnstr. 9, 64287 Darmstadt, GermanybInstitute for Atmosphere and Environment, Johann Wolfgang Goethe-University, Georg-Voigt-Street 14, 60325 Frankfurt, Germany

cIzana Atmospheric Observatory, La Marina 20, planta 6, Santa Cruz de Tenerife, 38071, Canary Islands, SpaindInstitute for Physics of the Atmosphere, Johannes-Gutenberg-University, Becherweg 21, 55099 Mainz, Germany

eDepartament of Geology, University of Huelva, Campus del Carmen, Huelva, 21071, Spain

Received 13 November 2006; received in revised form 21 June 2007; accepted 27 June 2007

Abstract

Samples from two strong homogeneous dust plumes from the Saharan desert reaching Izana (Tenerife, Spain) in July

and August 2005 were taken with a miniature impactor system and filter samplers. Size, aspect ratio and chemical

composition of more than 22,000 individual particles were studied by scanning electron microscopy. The mineralogical

phase composition of about 200 particles was investigated by transmission electron microscopy. In addition, the aerosol

size distribution was measured with an optical particle spectrometer. In all samples, the aerosol was dominated by mineral

dust with an average composition (by volume) of 64% silicates, 6% quartz, 5% calcium-rich particles, 14% sulfates,

1% hematite, 1% soot and 9% other carbonaceous material. Sulfate was found predominantly as coating on other

particles with an average thickness of approximately 60 nm. The aerosol calcium content is correlated with the calcite

concentrations of soils in the source region, highest values were observed for northern and central Algeria and Morocco.

The average aspect ratio of the particles was 1.64. The distributions of the aspect ratios are parameterized by log-normal

functions for modeling purpose. Single-scattering albedo (0.95) and asymmetry factor (0.74–0.81) was measured by polar

aerosol photometry on filter samples using a light source resembling the solar spectrum. The apparent soot content of the

sample (1 vol%) was determined by the same technique. From the mineralogical data, an average complex refractive index

of 1.59–9� 10�3i for visible light was derived. The imaginary part of the complex refractive index decreases with increasing

particle size from �2.5� 10�2i to o�10�3i, reflecting the decreasing hematite and soot contents. The imaginary part

derived from optical measurements was �7� 10�3i.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: Mineral dust; Electron microscopy; Chemical composition; Aspect ratio; Complex refractive index; Mixing state; Coatings

e front matter r 2007 Elsevier Ltd. All rights reserved.

mosenv.2007.06.047

ing author. Tel.: +49 615 116 4344; fax: +49 615 116 4021.

ess: kzk@gmx.de (K. Kandler).

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8059

1. Introduction

Mineral dust is produced over deserts by naturalprocesses and probably to a lesser extent byanthropogenic activities (Mahowald and Luo,2003; Mahowald et al., 2004; Tegen et al.,2004a, b). This airborne dust can be transportedover long distances (e.g., Prospero, 1999). The roleof desert aerosols in atmospheric processes stronglydepends on a variety of physicochemical parametersas well as the spatial distribution and transforma-tions during their atmospheric residence time. Dustoptical parameters, size and spatial distribution ofmineral aerosols determine the direct interaction ofsolar and terrestrial radiation in the atmospherethrough scattering and absorption (e.g., Liao andSeinfeld, 1998). Mineral dust also has a great impacton cloud formation including both ice and waterclouds leading to additional indirect radiationeffects. In addition, the bulk and surface chemistryof the mineral dust particles determines interactionswith gaseous and other particle species.

The global importance of desert dust with regardto the impact on the direct radiative forcing and itspossible effect on the indirect radiative forcingwas emphasized within the IPCC Report (2001).A direct forcing due to mineral dust on the order of70.5Wm�2 is discussed. However, quantificationof the magnitude of warming or cooling remainsopen because of the strong variability of theatmospheric dust burden and the lack of represen-tative data for the spatial and temporal distributionof the dust composition. Some basic processesgoverning production, transport, interaction withthe radiation field and removal from the atmosphereare generally understood. However, process simula-tions and model predictions strongly rely on over-simplifications and crude parameterizations of theseprocesses and of the physicochemical parametersinvolved (Sokolik et al., 2001). An improvement ofsuch model simulations requires refined source andemission data together with detailed data of the sizedistribution and the size-resolved chemical andmineralogical composition. For example, Sokolikand Toon (1999) have demonstrated how thevariation of mineral dust mixtures, causing differentrefractive indices, accounts for positive or negativeradiative forcings. In addition, desert dust cannotonly be regarded as a mixture of mineral compo-nents, because anthropogenic and marine air massesare frequently mixed to air masses dominatedby mineral dust (Kallos et al., 1998). Therefore,

heterogeneous reactions between the different com-ponents can be expected.

The spectral optical properties for mineral dustmixtures were determined by Sokolik and Toon(1999) for several desert regions, based on data setsof the refractive index of different mineral phasesand data of the major bulk composition. However,at present there are still not enough reliable dataavailable describing the size-resolved compositionof desert dust, both close to the source and aftertransformation during transport. The mineralogicaland chemical composition of dust were investigatedmostly on the basis of bulk samples (Glaccum andProspero, 1980; Schutz and Sebert, 1987; Stuutet al., 2005). In a few cases (Caquineau et al., 2002,1998), the clay fraction was investigated separately.Recently, individual particles analysis by electronmicroscopy was applied to Saharan dust (Blancoet al., 2003; Falkovich et al., 2001; Formenti et al.,2007; Hoornaert et al., 2003; Matsuki et al., 2007;Petit et al., 2005; Reid et al., 2003a).

Individual particle analysis combining scanningand transmission electron microscopy (SEM andTEM) can be used to derive size-resolved informa-tion of the composition, mixing state, morphology,and complex refractive index of atmospheric aerosolparticles (Ebert et al., 2004, 2002). Using automa-tion procedures, large numbers of individual parti-cles in the size range from 100 nm up to 50 mm indiameter can be studied, providing high statisticalsignificance. In the present paper, this individualparticle approach is applied to Saharan dust plumesat the Izana Observatory (Tenerife, Spain). Ouranalysis provides a very detailed data set on the size-resolved chemical composition and optical proper-ties for mineral dust.

2. Experimental

2.1. Sampling location and time

Aerosol samples were collected at Izana Observa-tory, which is located at 2367m above sea level(a.s.l.) on top of a ridge that crosses Tenerife islandSW to NE. This free troposphere site is above theinversion layer and stratocumulus-mantel located atE1 kma.s.l. in summer. It is linked to the synopticscale subsidence and trade winds blowing (Rodrıguezet al., 2004). In summer, the free tropospheresubsidence of NW air masses conditions prevailingat Izana often disappear, allowing the inflow of high

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748060

amounts of mineral dust from North Africa abovethe trade winds layer.

During the field campaign at Izana Observatory,we collected samples from two dust events. Thefirst occurred between 13 and 23 July 2005. In thefirst 2 days of this period PM10 raised from below10–80 mgm�3. The dust transport was prompted byhigh pressures over Algeria–Libya and a cyclone tothe NW of Tenerife leading to westward transportover the Western Sahara and Mauritania andsubsequent northward transport to Tenerife. Dur-ing 15–16 July 2005, western Atlantic ‘‘clean air’’conditions prevailed, reducing PM10 levels againbelow 10 mgm�3. During 17–23 July 2005, a mid-tropospheric anticyclonic circulation centered overMorocco–Algeria transported large amounts ofdust to Tenerife (PM10: 80–140 mgm�3). To illus-trate the situation, Fig. 1 shows the results of theDust Regional Atmospheric Model DREAM(Nickovic et al., 2001) for 20 July 2005. A dustplume extends over the Atlantic Ocean with itscenter south of the Canary Islands leading to acolumnar dust load between 1.5 and 4.5 gm�2 overTenerife. The back trajectory for Izana (black)points to an airmass origin in Algeria, whereasbelow the trade wind inversion Atlantic air stilldominated (red trajectory). Under meteorologicalconditions similar to those described above a second

Fig. 1. (a) Map of columnar dust load (gm�2) and 3000m wind

vectors for 20 July 2005 0-h forecast of DREAM model. Back-

trajectories (a and b) for the same date at 500 (red), 1500 (white)

and 2400 (black) meters are also shown.

Saharan dust episode took place during 6–8 August2005.

A contribution of marine boundary layer aerosolto the free-tropospheric Saharan dust plumes wasnot observed, in contrast to Hoornaert et al. (2003)who found significant amounts of Na and Cl atIzana. During both dust events investigated in thepresent paper a very stable trade wind inversionexisted. At the top of the marine boundary layertemperatures were below 20 1C, as compared to atemperature of more than 35 1C at the lower end ofthe free troposphere. This strong inversion clearlyinhibited any marine influence at the samplinglocation.

2.2. Backward trajectory analysis

Trajectory ensembles were calculated on the basisof 6-h operationally analyzed three-dimensionalwind fields from the European Centre for Med-ium-Range Weather Forecast (ECMWF) in 0.51horizontal resolution using the LAGRANTO soft-ware developed by Wernli and Davies (1997). Five-day backward trajectories were started at each gridpoint within the three-dimensional box 28–291N,16–171W, and 850–650 hPa.

2.3. Size distributions

Aerosol number size distributions were measuredin 15 size channels between 0.3 and 20 mm with anOptical Particle Counter (GRIMM model 1.108,Ainring, Germany). The measurements were vali-dated by comparing the particle volume calculatedfrom the number size distribution and the particlemass PM2.5 and PM10 deduced from gravimetry offilter samples.

2.4. Particle sampling

Sampling for single particle analysis was per-formed with a newly constructed single round jetcascade impactor, the Micro Inertial Impactor(MINI). The overall size of the stainless steelimpactor tube is 2 cm in diameter and 6 cm inlength. It can hold up to six individual stages. A setof nozzle inserts with an outer diameter of 1 cm andorifice diameters from 0.25 to 1.5mm were con-structed. The cone angle of the nozzle is 901. Thedistance from the nozzle to the impaction plate is setto 1.5 times the nozzle diameter. These nozzle sizesrepresent 50% efficiency cutoffs from just under

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8061

100 nm to 3 mm particle diameter with Reynoldsnumber ranging from 3500 to 500 at a sample flowof 8 cm3 s�1 (for cutoff calculation refer to Raabeet al., 1988). As the 0.25mm nozzle gets critical atthis sample flow, no flow regulation is needed. Theactual nozzles used for sampling had diameters of1.31, 0.69, 0.50, 0.35, and 0.25mm, which at Izanaconditions correspond to cutoff sizes of approxi-mately 2.5, 1.0, 0.5, 0.25, and 0.1 mm, respectively.

The small dimensions of the MINI allow mount-ing on a three-dimensional wind vane, whichprovides isoaxial collection of particles up to atleast 30 mm diameter. The wind vane was con-structed with an annular fin following the optimumdesign for propeller vanes originally intended forwind vector measurements (Monna and Driedonks,1979; Wieringa, 1967). A blunt inlet shapedaccording to NACA profile with a rounded inneredge (Baals et al., 1948) was used to minimizeparticle loss due to non-perfect isoaxial conditions(Hermann et al., 2001). Behind the inlet, a diffusorwith a cone angle of 71 was used to adapt theaverage wind speed to the sampling flow.

As the nozzle insert and the impaction substratemount are two separate pieces, different types ofsubstrates can be easily used within the MINI.Because of the bounce-off problem impactors sufferat the low humidity (e.g., Winkler, 1974) and theprevalent non-hygroscopic particles present inSaharan dust plumes, some adhesive must be usedto collect representative particle samples. As oil andgrease coatings (Turner and Hering, 1987) are notsuitable for electron microscopy, conductive andadhesive carbon tape was used as impaction surface.This carbon tape is well suited for mineral dustanalysis by SEM. For TEM, nickel TEM grids withformvar foil (Plano S-162 N3, Wetzlar, Germany)were used as sampling surface.

Due to the rather high aerosol concentrations ofthe investigated dust plumes very short samplingtimes between 20 s and 15min were sufficient.

2.5. Scanning electron microscopy

Approximately 22,000 particles were analyzed bySEM in two different instruments (Philips ESEMXL30 LaB6 and FEI ESEM Quanta 200 FEG,Eindhoven, The Netherlands). Most of the sampleswere investigated under high vacuum conditions of10�5 hPa, for selected samples low vacuum condi-tions of 1 hPa have been used. All samples werescreened for distinctive features before automated

particle analysis (software EDAX/AMETEK Gen-esis 4.52) was performed. Appropriate magnifica-tions were chosen for the different impactor stages:60 nm/pixel for the first stage, 30 and 15 nm/pixelfor the second and third stage, and 4 nm/pixel forthe smallest particles. Automated particle analysiswas performed on the carbon tape by use of thebackscatter electron signal, and on TEM grids byuse of the secondary electron signal. An accelerationvoltage of 20 kV, spot size 4, and a working distanceof 10mm were used for all automatically analyzedsamples.

Between 200 and 2000 individual particles wereinvestigated for each sample. The particle size isdetermined as equivalent projected area diameter.The aspect ratio AR is calculated by fittingan ellipse to the projected particle area and dividingthe major by the minor axis; Reid et al. (2003a)verified the reliability of this procedure. Theparticle volume was determined following theapproach of Anderson et al. (1996), who equalthe smallest particle dimension (from the projectedarea) with the particle height. Thus, the volume iscalculated as V ¼ p� d3/(6AR1/2). The chemicalcomposition of all elements with ZX6 of theparticles was determined in both instruments byenergy-dispersive X-ray microanalysis with Si(Li)detectors. X-ray count rates were corrected formatrix effects using the so-called ‘‘standard-less’’procedures provided by the EDAX Genesissoftware. Due to the large number of particlesinvestigated, it was not possible to apply correctionprocedures for geometric effects (Armstrong, 1991).Therefore, the resulting atomic composition mustbe regarded as a rough estimate of the chemicalcomposition of the particles. According to Arm-strong (1991), neglecting geometric effects may leadto large errors in the chemical composition up to30% (relative).

2.6. Transmission electron microscopy

TEM was performed with a Philips CM20instrument (LaB6 cathode) using an accelerationvoltage of 200 kV. The TEM is equipped with anenergy-dispersive X-ray detector with a high-purityGe crystal (Thermo NORAN, Middleton, USA)allowing detection of all elements with ZX6(carbon). Phase identification is based on electrondiffraction patterns which were indexed using theICSD database (FIZ Karlsruhe, 2006) and thesoftware package PIEP (Miehe, 1997).

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748062

Phase determination of individual particles byanalyzing electron diffraction patterns is time-consuming. Therefore, this technique was onlyapplied to a small number of particles (about 200)within the silicate class. For most other classes, thephase composition was deduced from the chemicalcomposition assuming the presence of the mostabundant natural phase. For example, particlescontaining Ca, S and O as major elements wereassumed to be gypsum (CaSO4� 2H2O) and notanhydride (CaSO4).

2.7. Polar aerosol photometry

Aerosol samples for polar photometry werecollected during the dust events on Nucleporemembrane filters (0.2mm pore size; Corning Costa,Tokyo, Japan) with a sampling time of approxi-mately 1 day. The surface of the filter is coveredhomogeneously with aerosol particles forming anoptical two layer system. This system can bedescribed by an exact radiation transfer model(Bundke et al., 2002; Hanel, 1994). The lightscattered by and transmitted through the filter ismeasured in a polar aerosol photometer, both beforeand after loading the filter with particles. Thehemispherical integrals of the forward-scattered andtransmitted radiation and of the backward-scatteredradiation are calculated, as well as an upper limit ofthe transmittance. The optical properties of theparticles are derived from these properties using theinversion algorithm of Hanel (1994).

Primarily, the optical depth of the particle layer t,the single scattering albedo o, and the asymmetryparameter g of the volume scattering function arederived using the Henyey–Greenstein function withan error below 5%. To address the soot content, amixture of non-soot mineral dust particles (index‘‘c’’ in the following equations) and soot (index ‘‘s’’)is assumed. The refractive index of the non-sootparticles at 700 nm (1.557–0.002i) is calculated fromthe model mineralogy derived from the singleparticle measurements (see below), for soot a valueof 1.76–0.45i was used. The apparent volumefraction of soot fs is derived from the singlescattering albedo, using a k ¼ k(o) relation andthe Lorentz–Lorenz refractive index mixing rule fortwo components (Bundke et al., 2002; Hanel, 1988):

n ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3þ kx

1þ ðk2=5Þ � zÞ

s,

x ¼Rðnc; kcÞ � Rðns; ksÞ

Aðns; ksÞ � Aðns; ksÞ,

z ¼ Rðnc; kcÞ � Aðns; ksÞx,

Rðn; kÞ ¼n2 � 1

n2 þ 2þ

k2

5,

Aðn; kÞ ¼k

n2 þ 2,

f s ¼Aðn; kÞ � Aðnc; kcÞ

Aðns; ksÞ � Aðnc; kcÞ.

The error of our approach mainly depends on theassumptions used for the calculations. The opticalproperties apply to the dry aerosol, since photo-metric measurements were performed at relativehumidities below 30%. All results are solar spectralmean values. According to Blanchet (1982), they arecomparable to spectral values at the wavelengthl ¼ 700 nm.

3. Results and discussion

3.1. Chemical and mineralogical composition

3.1.1. Particle classes

Size (equivalent projected area diameter), aspectratio and chemical composition of more than 22,000particles from 39 samples were measured. Based onthe chemical composition derived from X-ray countrates, the particles were classified into 11 differentclasses (criteria are given in Table 1). More than99.5% of all particles could be classified into one ofthe 11 classes. The rest of the particles is summar-ized in the class ‘‘other’’ and is neglected in thefurther analysis. Mineral phases observed byselected area electron diffraction in TEM andminerals determined indirectly from chemical com-position are also given in Table 1.

Typical secondary electron images of particles ofthe silicates, carbonates, and gypsum class areshown in Fig. 2. Please note the heterogeneouscomposition of the particles. Small silicate grainsoccur in most particle groups (two examples areshown in Fig. 2a and b). In addition, small grains ofiron oxides occur on the surface of silicates (Fig. 2c).One rare example is shown in Fig. 2d, where a rareearth element phosphate is visible on the surface ofa silicate. Bright-field images and electron diffrac-tion patterns of particles from the silicate class (clayminerals and anorthite) are displayed in Fig. 3.

ARTICLE IN PRESS

Table 1

Definition of particle classes, phases determined by electron diffraction in TEM and derived from chemical composition

Class name Criteria Phases identified by

TEM

Phases deduced from

particle chemistry

Iron-rich jFej40.15 and Si/Feo1 and Ti/Feo1.33 Hematite Ilmenite

Titanium-rich jTij40.3 and Na/Tio1 and Mg/Tio1 and Al/Tio1

and Si/Tio1 and S/Tio1 and Fe/Tio1

Rutile

Carbonates [jCa+Mgj40.5 and 0.33oMg/Cao3] or [jCaj40.5

and Mg/Cao0.33 and Si/Cao0.5 and S/Cao0.25

and P/Cao0.15]

Dolomite Calcite

Other calcium-rich jCaj40.25 and Si/Cao0.5 and not in class

carbonates or gypsum

Apatite

Gypsum jCa+Sj40.5 and 0.25oCa/So4 and Na/Cao0.5 Gypsum

Halite jNa+Clj40.5 and S/Nao0.375 and S/Clo0.5 and

Si/Clo0.5

Halite

Quartz jSij40.5 and Mg/Sio0.2 and Al/Sio0.2 Quartz

Silicates jSij40.2 and Na/Sio0.7 and Mg/Sio1.33 and Al/

Sio1.33 and K/Sio0.5 and Ca/Sio0.5 and Ti/

Sio0.5 and Fe/Sio0.5 and jP+S+Cljo0.2 and not

in class quartz

Biotite, kaolinite,

illite, muscovite,

chloritoide, anorthite

Chlorite, albite, leucite,

orthoclase, plagioclase,

chrysotile

Sulfate silicate

mixtures

jNa+S+Mg+Al+Si+K+Caj40.7 and 0.6oS/

Sio2

Any silicate and sulfate

Sulfates jSj40.5 and Cl/So0.3 and Si/So0.5 and Fe/So0.5

and not in class gypsum

Blodite/loweite, glauberite,

vanthoffite, mirabilite/

thenardite, mascagnite

Carbonaceous JC+OJ40.97

Other Not in other class Alunite, periclase/

magnesite, spinel

All elemental ratios based on moles; jXj ¼ total atomic content of element X relative to all other elements analyzed (except C and O); JXJ

total atomic content of element X relative to all other elements analyzed (including C and O).

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8063

It is important to note here, that we haveclassified all carbon-rich particles as carbonaceousparticles. As we have used in most cases a carbontape as substrate, it was impossible to detect soot. Ingeneral, soot can be easily recognized in SEM due toits characteristic morphology (Ebert et al., 2004,2002). However, the surface morphology of thecarbon tape is similar to the morphology of sootagglomerates. Therefore, the fraction of soot withinthe class of carbonaceous particles had to beestimated from polar aerosol photometry. Thistechnique yielded a soot fraction of approximately10% within the class of carbonaceous particles.

The average atomic composition of the elevenclasses, as well as the total composition, is summarizedin Table 2. The carbonaceous class is not includedbecause these particles consist mainly of the elementsC and O, which were not quantified in our analysisdue to the influence of the substrate. Some classeshave a rather narrow composition which is close to apure mineral (e.g., gypsum or quartz), while otherclasses represent mixtures of different phases and,therefore, exhibit a large compositional variation.

The variable composition of many particle classescan be visualized with the help of an element index.This index is defined as the atomic ratio of theconcentration of the element considered and thesum of the concentrations of the elements Na, Mg,Al, Si, P, S, Cl, K, Ca, Ti, Cr, Mn, Fe, Co, and Cu.The concentration of C and O in the particlescannot be determined accurately due to a contribu-tion of the sampling substrate. Therefore, they werenot included into the calculation of the elementindex. Besides these major and minor elementsrepresenting almost the total mineral dust composi-tion, a few trace components (Sr, Zr, Ba, La, Ce,Nd) were detected, but were ignored due to theirrare or unique occurrence. The aluminum andsilicon indices are shown in Fig. 4 for the silicateand quartz classes, as well as for all other particles.The location of some pure minerals is also shown.Three important conclusions can be drawn fromFig. 4: First, the aluminum and silicon indices showa large absolute variation on the order of 0.2–0.3 inall classes. Second, the silicates class and to asmaller extent the quartz class consist of a mixture

ARTICLE IN PRESS

Fig. 2. Secondary electron images and X-ray spectra of (a) gypsum, (b) calcite, (c) and (d) silicates with highlighted grains of high iron and

rare earth element contents.

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748064

of several minerals. Third, the aluminum to siliconratio observed for the silicates class is identical tothe value observed for the other classes. This clearlydemonstrates that the same silicate minerals dom-inating the silicates class were mixed with thematerials dominating the other classes.

The size-resolved relative number abundance ofthe different particle classes averaged for both dustplumes is shown in Fig. 5. Above 1 mm particle size,the aerosol is dominated by mineral dust. From1 mm down to 100 nm diameter, the relative fractionof mineral dust continuously decreases from 490%

ARTICLE IN PRESS

Table 2

Average relative composition (atomic percent) of the particle classes

Class name Na Mg Al Si P S Cl K Ca Ti Cr Mn Fe Co Cu

Iron-rich 4 3 12 16 2 3 1 1 2 4 2 1 48 1 1

Titanium-rich 2 2 9 14 1 2 1 1 1 59 1 1 5 1 1

Carbonates 3 9 6 10 1 2 1 1 62 1 1 1 2 1 1

Other calcium-rich 4 6 12 24 3 3 1 2 35 1 1 1 4 1 1

Gypsum 4 2 3 6 2 38 1 1 37 1 1 1 2 1 1

Halite 47 3 2 2 1 3 38 1 1 0 0 0 1 0 0

Quartz 2 2 6 81 2 1 0 1 1 1 0 0 2 0 0

Silicates 4 5 26 46 2 2 1 3 2 1 0 1 6 1 0

Sulfate silicate mixtures 14 6 15 20 4 14 6 3 5 2 2 2 5 2 1

Sulfates 31 8 5 5 6 24 3 3 6 2 2 2 2 2 1

Total 5 5 21 45 2 3 1 3 6 1 1 1 6 1 0

Fig. 3. TEM bright field images and electron diffraction patterns of two typical silicate particles: (a) clay particle (kaolinite and

muscovite), and (b) feldspar (anorthite).

Fig. 4. Aluminum index versus silicon index for the silicates and

quartz classes and for all other particles. Indices of several pure

minerals are also shown.

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8065

to �40% (including half of the silicate sulfatemixture class). The relative contribution of sulfatesand carbonaceous material increases with decreas-ing particle size. The values for particles withdiameters o250 nm have a high uncertainty dueto small number of particles observed and dueto the weak X-ray signal obtained from theseparticles.

The overall mass composition is dominated bymineral dust. Using a mineralogical model compo-sition (see below) and the measured aerosol sizedistribution, the following bulk composition (byvolume) was obtained: 64% silicates, 6% quartz,5% calcium-rich particles, 14% sulfates, 1% hema-tite, 1% soot and 9% other carbonaceous material.

ARTICLE IN PRESS

0

10

20

30

40

no

rma

lize

d e

lem

en

t in

de

x (

ele

me

nt

ind

ex ×

pa

rtic

le v

olu

me

/ p

art

icle

su

rfa

ce

are

a)

0 2 4 6 8 10

particle diameter, µm

sulfur

calcium

iron

silicon

Fig. 6. Normalized element index versus particle diameter for

sulfur, calcium, iron, and silicon. For each element, all data are

divided by the value of the particle with smallest diameter.

Fig. 5. Size-resolved relative number abundance of the different particle groups averaged for both dust plumes. The total number of

particles (n) analyzed in each size bin is also given.

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748066

The relative abundances of quartz, calcite anddolomite observed at Izana are roughly half of thevalues found over Spain (Avila et al., 1996), CapeVerde (Glaccum and Prospero, 1980), and estima-tions on the long range transport by Schutz (1989).Instead, the relative silicate abundance is higher.

The average elemental mass ratios in this work—2.14 (Si/Al), 0.20 (K/Al), 0.28 (Ca/Al), and 0.58(Fe/Al)—are very similar to values reported byFormenti et al. (2003) for the Cape Verde Islands(2.18, 0.21, 0.35, and 0.53, respectively). Measure-ments of dust-containing rain samples over south-ern Italy (Blanco et al., 2003) yielded similar resultsfor Si/Al, K/Al and Ca/Al ratios (2.18, 0.30, 0.26),but a significantly lower Fe/Al ratio (0.22). Reidet al. (2003a) reported values for Saharan dust atPuerto Rico of 2.13 (Si/Al) and 0.17 (K/Al), but ahigher Ca/Al ratio of 0.38 and a lower Fe/Al ratioof 0.30. Most probably, this difference is relatedto different source regions. In general, the averageSi/Al and K/Al ratios of Saharan mineral dusttransported into northerly and westerly directionsseem to be rather constant, whereas the Ca/Al andFe/Al ratios show a higher variability.

3.1.2. Sulfate coatings

The distribution of the various element indicescalculated for all particles provides more detailedinsights into the size dependence of the elementalcomposition. The sulfur index is high (0.1–0.3) forsmall particles below 1 mm diameter, and decreaseswith increasing size to zero at 20 mm particlediameter. We have defined a normalized elementindex (element index�particle volume/particle sur-face area) in order to study the element distribution

within the particles. In Fig. 6, the normalizedelement index is plotted versus the particle diameter.In such a graph it is easy to distinguish betweenelements that occur within surface coatings andelements that are contained in the complete particlevolume, assuming that a coating thickness isindependent of particle size. Sulfur shows a constantnormalized element index (Fig. 6) indicating itspresence in a surface coating. In contrast, theelements calcium, iron, and silicon show a linearincrease of the normalized element index withparticle size, indicating their presence in thecomplete particle volume. Gypsum particles areexcluded from this analysis, as they contain sulfur intheir whole volume.

ARTICLE IN PRESS

Table 3

Mineralogical model composition (volume percent) and density

of the different particle classes

Class name Approx.

composition, vol%

Average density,

g cm�3

Iron-rich 40% hematite 4.1

30% illite

10% ilmenite

10% chloritoid

10% albite

Titanium-rich 60% rutile 3.8

20% illite

10% chloritoid

10% albite

Carbonates 75% calcite 2.8

15% dolomite

10% illite

Other calcium-

rich

50% calcite 2.8

20% illite

10% dolomite

10% chloritoide

10% albite

Gypsum 90% gypsum 2.3

10% blodite

Halite 80% halite 2.2

20% blodite

Quartz 85% quartz 2.6

15% illite

Silicates 55% illite 2.9

25% chloritoid

20% albite

Silicate sulfate

mixtures

30% illite 2.7

30% blodite

20% glauberite

10% chloritoid

10% albite

Sulfates 60% blodite 2.5

30% glauberite

10% illite

Carbonaceous 90% carbonaceous 1.5

10% soot

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8067

It appears from our data that most mineral dustparticles have a coating of sulfates. Possible originsof this coating include the uptake of sulfateprecursor gases on non-reacting surfaces in theabsence of water (Bauer and Koch, 2005; Usheret al., 2003) and cloud processing (Levin et al., 2005;Wurzler et al., 2000). However, due to the very dryconditions the aerosol encountered during trans-port, the first process is more probable in our case.It should be mentioned here, that the occurrence ofsulfate coatings on mineral dust was detected for theeastern Mediterranean with a similar reasoning(Levin and Ganor, 1996).

In order to obtain a rough estimate of thethickness of these coatings, a binary systemconsisting of a silicate sphere and a sulfate coatingwith a thickness independent of the core size isassumed. The volume fraction of the two materialsis then calculated from the silicon and sulfateindices; only data from the silicates, silicate sulfatemixture, and sulfates particle classes were used. Forthese calculations, the average molar weights anddensities were derived from the mineralogical modeldata (see below). An average coating thickness of60 nm sulfate on a silicate core was obtained. Itshould be emphasized here that this value has alarge uncertainty of factor 2–3. Despite these largeuncertainties, detection of small surface coatingsand quantification of the layer thickness on a verylarge number of particles (more than 16,000) bySEM and X-ray microanalysis is from a methodo-logical point of view impressive.

3.2. Complex refractive index and optical properties

For calculation of the complex refractive indexfor visible light, a mixing rule has to be applied toderive the average refractive index for each particlesize interval from the bulk refractive indices of thepure mineral phases and the relative dust miner-alogical composition determined by single particleanalysis. As detailed information on the internalparticle structure is available for few particles only,a simple volume mixing rule is applied to the data.This approach proofed to be successful in a closureexperiment on the optical properties of ruralaerosols in Central Europe (Ebert et al., 2002).For the size-resolved complex refractive index,knowledge of the size distribution is not necessary.Calculation of this material property can beperformed directly with the chemical compositionobtained from electron microscopy. In contrast,

calculation of the complex refractive index of thetotal aerosol requires the particle volume distribu-tion, which was measured with an optical particlecounter.

The volume fraction of the different mineralphases was estimated from a mineralogical modelcomposition (Table 3), which describes the measuredchemical composition best. The mineralogical phasesassigned to the different particle classes were taken

ARTICLE IN PRESS

Table 4

Average mineralogical model composition (volume percent) of the various size ranges

Diameter range, mm 0.05–0.1 0.1–0.25 0.25–0.5 0.5–1 1–2.5 2.5–5 5–10 10–20 Total 40.3mm

Illite 22 29 28 36 41 45 47 45 37

Albite 7 10 9 13 14 16 17 16 12

Chloritoid 8 11 11 15 18 19 21 19 15

Quartz 1 3 5 8 9 7 10 6

Calcite 2 5 6 5 6 3

Dolomite 1 1 1 1 1

Gypsum 1 1 1 1 1 1 1

Halite 1 1 1

Blodite 24 15 19 9 3 1 1 9

Glauberite 13 8 10 5 2 5

Hematite 2 5 1 1 1 1 1

Ilmenite 1

Rutile 2

Carbonaceous 25 18 17 12 4 1 9

Table 5

Complex refractive indices for visible light

Material Refractive index

Illite 1.57a

Albite 1.53a

Chloritoid 1.73a

Quartz 1.55a

Calcite 1.60a

Dolomite 1.62a

Gypsum 1.52a

Halite 1.54a

Blodite 1.49a

Glauberite 1.52a

Hematite 3.05–0.3ib

Ilmenite 2.4–0.3ia,c

Rutile 2.76a

Carbonaceous 1.5d

Soot 1.5–0.47id

aBarthelmy (2006).bChen and Cahan (1981), 590 nm.cImaginary part assumed to be equal to hematite, as the

reflectance of ilmenite is similar to hematite.dHorvath (1993).

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748068

from our TEM observations and from literature dataof the most common minerals observed in desert dust(Avila et al., 1997; Coude-Gaussen et al., 1987;Glaccum and Prospero, 1980; Molinaroli, 1996). Thechemistry of phases with highly variable composition(e.g., clay minerals) was taken from Barthelmy(2006), who lists the most common compositions ofthese phases. The average density of the differentmineral classes was also derived from the miner-alogical model composition. The soot content wasestimated by assuming a constant ratio between sootand carbonaceous material of 1:9. This ratio wasdetermined by the measured bulk soot content frompolar aerosol photometry and the volume fraction ofthe carbonaceous material derived from singleparticle analysis.

The size-resolved composition (volume percent)and the volume-weighted total composition (forparticles with diameter 40.3 mm) of mineral dustarriving at Izana has been calculated (Table 4). Toderive the overall average composition, the aerosolvolume distribution calculated from the opticalparticle counter was used.

The size-resolved complex refractive index forvisible light was calculated by using a volumemixing rule and the refractive indices of the puremineral phases (Table 5). The calculated complexrefractive index of the total mineral dust arriving atIzana was 1.59–9� 10�3i. The polar photometrymeasurements lead to a value of the imaginary partof �7� 10�3i for 700 nm wavelength. Both valuesare in excellent agreement with direct opticalmeasurements of Patterson et al. (1977), who reporta refractive index of approximately 1.56–7� 10�3i

at 550 nm wavelength and �4� 10�3i at 700 nm foralmost the same location (Pico de Teide, Tenerife,Spain). For the Cape Verde Islands, Carlson andCaverly (1977) found marginally lower values of theimaginary part of �2.9� 10�3i at 610 nm and�8� 10�3i at 468 nm.

The complex refractive index at Izana (Fig. 7)shows high values for both, the real part and theimaginary part for the smallest particles. The real partis slightly decreasing with increasing particle size,reflecting the decreasing abundance of iron oxides.

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8069

Absorption is strongly increasing with decreasingparticle size due to the higher abundance of soot andiron oxides within the small particle size range.

For solar wavelengths, the average (bulk) extinc-tion coefficient–derived from polar aerosol photo-metry–was 75Mm�1, consisting of a scatteringcoefficient of 71Mm�1 and an absorption coeffi-cient of 3.8Mm�1. The resulting single-scattering

1.50

1.55

1.60

1.65

1.70

refr

active index,

real part

(n)

0.000

0.005

0.010

0.015

0.020

0.025

0.030

refr

active index, im

agin

ary

part

(k)

0.1 1.0 10.0

particle diameter, µm

n

k

Fig. 7. Size-resolved complex refractive index n–ki (visible light)

derived from single particle analysis.

0.0

0.2

0.4

0.6

0.8

1.0

rela

tivem

assf

raction

0.0

0.2

0.4

0.6

0.8

1.0

rela

tive

mass

fraction

14.7 15.7 16.7 17.7 18.7 19.7

Fig. 8. Time series of relative mass abundance of the different particl

1mmodo2.5mm (lower part) for the dust plume reaching Izana in Jul

albedo for all samples varies between 0.945 and0.955 and is in perfect agreement with measure-ments over the Southern Sahara (0.95) reported byFouquart et al. (1987). The Henyey–Greensteinasymmetry parameter varies between 0.74 and 0.81for the Saharan dust, which resembles the valuesof 0.75–0.78 (610 nm wavelength) found on theCape Verde Islands (Carlson and Caverly, 1977)and those calculated for model Saharan dust of0.75–0.81 (Kalashnikova and Sokolik, 2002). How-ever, except for a very slight decrease in asymmetryparameter with increasing calcium index, no depen-dency on the chemical composition was found.

3.3. Variability of chemical composition and source

region

Time series of the relative mass fraction of thedifferent particle groups are shown in Fig. 8 for thetwo size ranges 1 mmodo2.5 mm and 2.5 mmodo10 mm. The time series represent the first dustplume which arrived at Izana in July 2005.Composition of the plume appears to be rather

20.7 21.7 22.7

iron-rich

titanium-rich

carbonates

other calcium-rich

gypsum

halite

quartz

silicates

mixtures

sulfates

carbonaceous

other

e groups for the size ranges 2.5 mmodo10 mm (upper part) and

y 2005.

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748070

homogeneous, except for the increase of thecalcium-containing particles in the size range of2.5 mmodo10 mm after 19 July, which is probablyrelated to a change in source region (see below). Thesmaller particles show a higher variability than thelarger ones. This observation cannot be explainedby the counting statistics, as a large number ofparticles were investigated in both size ranges.

The difference in composition between the twoinvestigated dust plumes is also low. For example,the quartz mass fraction increases from 9% to 11%from the July plume to the August plume, whereasthe silicate fraction decreases from 80% to 77%.Both differences are not regarded as significant. Thesmall variations of the chemical composition supportthe low variability found for the optical parameters.

The possible source regions of the samples weredetermined from 120 h backward trajectories en-sembles for each sampling hour. Trajectory coordi-nates (time step of 1 h) with an altitude below1000ma.s.l. are displayed in Fig. 9 in order torepresent locations of potential dust uptake. Moun-tain regions are excluded by this method, but are

0.04 0.06 0.08

calciu

15°N

20°N

25°N

30°N

35°N

40°N

45°N

latitu

de

20°W 15°W 10°W 5°W

lon

Fig. 9. Potential dust uptake regions derived from backward tr

known to produce only small amounts of dust(Prospero et al., 2002). The different locations arecolored according to the calcium index of therelated sample for particles with 1 mmodo10 mm.

It can be clearly seen from Fig. 9 that the calciumindex is strongly decreasing from north to south.This is in good agreement with soil maps given byDesboeufs and Cautenet (2005) and the work ofSchutz and Sebert (1987), which show high calciteconcentrations for northern and central Algeria andMorocco, and low concentrations south of approxi-mately 27 1N. Even the low calcite content insouthern Tunisia and the high values of coastalLibya are reproduced quite well. Only an isolatedregion of higher calcite content in southern Algeriadescribed by Desboeufs and Cautenet (2005) wasnot observed in our data. However, it has to be keptin mind that only one calcium index value isavailable as an integral over the whole trajectorypath, so isolated sources may not be recognizedvery well. The iron index map (not displayed)shows approximately the opposite tendency, asthe sample-averaged calcium and iron index are

0.10 0.12 0.14

m index

0° 5°E 10°E 15°E 20°E

gitude

ajectories, colored according to the aerosol calcium index.

ARTICLE IN PRESS

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

de

nsity

1 2 3 4 5 6 7 8

aspect ratio

Fig. 10. Measured and parameterized (solid line) aspect ratio

density distribution for the silicates class.

1

2

3

4

5aspect ra

tio

50 to 100 particles

> 100 to 500 particles

> 500 to 1000 particles

> 1000 to 5000 particles

0.03 0.05 0.07

sulfur index

0.09 0.11 0.13 0.15

Fig. 11. Relation between aspect ratio and sulfur index. Ranges

for the number of particles analyzed are indicated by different

symbols.

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8071

anticorrelated with a correlation coefficient of�0.71. This anticorrelation reflects the fact thatiron is associated mainly with silicates—especiallywith the clay minerals—and is not found in calcium-rich particles (carbonates and gypsum).

Due to the prevailing zonally-orientated trajec-tories, a potentially existing east–west gradientcannot be detected. It is important to note thatalmost no potential source region is located over theAtlantic Ocean or the Mediterranean Sea. There-fore, halite observed in the samples most likelyoriginates from salt lakes within the Sahara.

3.4. Aspect ratios of individual particles

The aspect ratio (AR) density distribution foreach particle class was determined. The densitydistributions are parameterized (parameters can befound in Table 6) with a modified lognormalfunction h(AR):

hðARÞ ¼1ffiffiffiffiffiffi

2pp

sðAR� 1Þ

� exp �1

2

lnðAR� 1Þ � ms

� �2 !

.

As an example, Fig. 10 shows the parameteriza-tion curve together with the measured values for thesilicates particle class.

The median for each particle class is given inTable 6. Halite has the lowest value (1.44) reflectingthe cubic particle morphology. The highest value(1.92) was observed for sulfates indicating theirprevalent needle-like geometry. The median aspectratio of all particles (1.64) is significantly lower thanthe value of E1.9 reported by Reid et al. (2003a)

Table 6

Parameters of aspect ratio density distribution and median aspect

ratio for different particle classes

Particle class s m Median

Iron-rich 0.5845 �0.5475 1.59

Titanium-rich 0.5734 �0.6289 1.54

Carbonates 0.6188 �0.4764 1.62

Other calcium-rich 0.6185 �0.4331 1.67

Gypsum 0.7484 �0.2810 1.77

Halite 0.6520 �0.9359 1.44

Quartz 0.6241 �0.5218 1.60

Silicates 0.6508 �0.4820 1.62

Sulfate silicate mix. 0.6279 �0.1818 1.83

Sulfates 0.7309 �0.1104 1.92

Total 0.6579 �0.4502 1.64

and Reid et al. (2003b). For sulfates, the relationbetween aspect ratio and chemical composition isshown in Fig. 11. Aspect ratio and sulfate concen-trations are positively correlated, but particles withextreme aspect ratios are rare.

The size dependence of the median aspect ratio aswell as its variability (0.1–0.9 quantiles) is shown inFig. 12. There is a clear minimum of the aspect ratioaround 3 mm particle diameter. The increase of theaspect ratio towards smaller particles is mostprobably linked to their higher sulfur contents.

For modeling purposes, size-resolved aspect ratiodensity distributions can be calculated from thevalues given in Table 7 using the function h(AR), asnoted above. Below 250 nm, however, the statistical

ARTICLE IN PRESS

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.62.83.0

3.3

3.6

aspect ra

tio

1 10

particle diameter, µm

Fig. 12. Size dependence of the median aspect ratio and the 0.1

and 0.9 quantiles (shaded area).

Table 7

Parameters of aspect ratio density distribution and median aspect

ratio for particle size intervals

Diameter range, mm s m Median

0.06–0.1 0.5945 0.2569 2.26

0.1–0.25 0.5012 0.1289 2.09

0.25–0.5 0.6499 �0.0189 1.99

0.5–1 0.6256 �0.2914 1.75

1–2.5 0.6555 �0.4815 1.62

2.5–5 0.6456 �0.5702 1.57

5–10 0.6256 �0.3815 1.68

10–20 0.6096 �0.2959 1.76

K. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748072

spread is very high and the number of particlesanalyzed is low, leading to a high uncertainty.

4. Conclusions

1.

The composition of the two dust plumes arrivingat Izana (Tenerife, Spain) in July and August2005 is quite similar to previous dust outbreaksat the same location, the Cape Verde Islands andSouthwest Europe. High calcium contents inthe aerosol point to a source region mainlyin northern and central Algeria as well as inMorocco.

2.

Above 1 mm particle diameter, the aerosol sam-ples are dominated by mineral dust (more than90% by number). With decreasing particle size,sulfates and carbonaceous material become more

important and dominate the aerosol compositionbelow 500 nm. Below 100 nm particle diameter,the number abundance of silicates is approxi-mately 40%.

3.

Sulfates were found to occur predominantly ascoating on other particles with an averagethickness of approximately 60 nm. The majorityof particles are coated.

4.

An average complex refractive index of1.59–7� 10�3i was obtained from the mineralo-gical model composition derived by electronmicroscopy. This value is in excellent agreementwith direct optical measurements at Tenerife andthe Cape Verde Islands. The imaginary partstrongly decreases with increasing particle size,reflecting the fact that the highly absorbingcomponents (hematite and soot) are predomi-nantly found in the small particle range.

Acknowledgments

This work is part of the SAMUM research group.We would like to thank all members of theSAMUM consortium for the stimulating coopera-tion. Financial support by the Deutsche For-schungsgemeinschaft is gratefully acknowledged.We also want to thank the Barcelona Supercomput-ing Centre for providing forecasts with theDREAM model (http://bscct01.bsc.es/DREAM/)which were used during the field campaign and fordata analysis. Special thanks go to Dr. G. Miehe(TU Darmstadt) for his help in evaluating theSAED patterns.

References

Anderson, J.R., Buseck, P.R., Patterson, T.L., Arimoto, R.,

1996. Characterization of the Bermuda tropospheric aerosol

by combined individual-particle and bulk-aerosol analysis.

Atmospheric Environment 30 (2), 319–338.

Armstrong, J.T., 1991. Quantitative elemental analysis of

individual microparticles with electron beam instruments.

In: Heinrich, K.F.J., Newbury, D.E. (Eds.), Electron Probe

Quantitation. Plenum Press, New York, pp. 261–315.

Avila, A., Queralt, I., Gallart, F., Martin-Vide, J., 1996. African

dust over northeastern Spain: Mineralogy and source regions.

In: Guerzoni, S., Chester, R. (Eds.), The Impact of Desert

Dust Across the Mediterranean. Kluwer Academic Publish-

ers, Dordrecht, pp. 201–205.

Avila, A., Queralt-Mitjans, I., Alarcon, M., 1997. Mineralogical

composition of African dust delivered by red rains over

northeastern Spain. Journal of Geophysical Research 102

(D18), 21977–21996.

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–8074 8073

Baals, D.D., Smith, N.F., Wright, J.B., 1948. The development

and application of high-critical-speed nose inlets. NACA

Report 920, 535–592.

Barthelmy, D., 2006. Mineralogy Database. /http://www.

webmineral.comS, accessed on 16 October 2006.

Bauer, S.E., Koch, D., 2005. Impact of heterogeneous sulfate

formation at mineral dust surfaces on aerosol loads and

radiative forcing in the Goddard Institute for Space Studies

general circulation model. Journal of Geophysical Research

110, D17202.

Blanchet, J.-P., 1982. Application of the Chandrasekhar mean to

aerosol optical parameters. Atmosphere—Ocean 20, 189–206.

Blanco, A., et al., 2003. Characterization of African dust over

southern Italy. Atmospheric Chemistry and Physics 3,

2147–2159.

Bundke, U., et al., 2002. Aerosol optical properties during the

Lindenberg aerosol characterization experiment (LACE 98).

Journal of Geophysical Research 107 (D21), 8123.

Caquineau, S., Gaudichet, A., Gomes, L., Magonthier, M.-C.,

Chatenet, B., 1998. Saharan dust: clay ratio as a relevant

tracer to assess the origin of soil-derived aerosols. Geophy-

sical Research Letters 25 (7), 983–986.

Caquineau, S., Gaudichet, A., Gomes, L., Legrand, M., 2002.

Mineralogy of Saharan dust transported over northwestern

tropical Atlantic Ocean in relation to source regions. Journal

of Geophysical Research 107 (D15), 4251.

Carlson, T.N., Caverly, R.S., 1977. Radiative characteristics of

Saharan dust at solar wavelengths. Journal of Geophysical

Research 82 (21), 3141–3152.

Chen, C.T., Cahan, B.D., 1981. Visible and ultraviolet optical

properties of single-crystal and polycrystalline hematite

measured by spectroscopic ellipsometry. Journal of the

Optical Society of America 71 (8), 932–934.

Coude-Gaussen, G., et al., 1987. Saharan dust on Fuerteventura

island (Canaries): chemical and mineralogical characteristics,

air mass trajectories, and probable sources. Journal of

Geophysical Research 92 (D8), 9753–9771.

Desboeufs, K.V., Cautenet, G., 2005. Transport and mixing zone

of desert dust and sulphate over Tropical Africa and the

Atlantic Ocean region. Atmospheric Chemistry and Physics

Discussion 5, 5615–5644.

Ebert, M., et al., 2002. Complex refractive index of aerosols

during LACE 98 as derived from the analysis of individual

particles. Journal of Geophysical Research 107 (D21), 8121.

Ebert, M., Weinbruch, S., Hoffmann, P., Ortner, H.M., 2004.

The chemical composition and complex refractive index of

rural and urban influenced aerosols determined by individual

particle analysis. Atmospheric Environment 38, 6531–6545.

Falkovich, A.H., Ganor, E., Levin, Z., Formenti, P., Rudich, Y.,

2001. Chemical and mineralogical analysis of individual

mineral dust particles. Journal of Geophysical Research 106

(D16), 18029–18036.

FIZ Karlsruhe, 2006. Inorganic Crystal Structure Database

(ICSD). /http://www.fiz-karlsruhe.de/ecid/Internet/en/DB/

icsd/S, accessed on 22 September 2006.

Formenti, P., Elbert, W., Maenhaut, W., Haywood, J., Andreae,

M.O., 2003. Chemical composition of mineral dust aerosol

during the Saharan dust experiment (SHADE) airborne

campaign in the Cape Verde region, September 2000. Journal

of Geophysical Research 108 (D18), 8576.

Formenti, P., et al., 2007. Size distribution, morphology, and

composition of mineral dust and biomass burning aerosols

from western Africa as observed by scanning and transmis-

sion electron microscopy. Geophysical Research Abstracts 9,

09185.

Fouquart, Y., et al., 1987. Observations of Saharan aerosols:

results of ECLATS field experiment, Part II: broadband

radiative characteristics of the aerosols and vertical radiative

flux divergence. Journal of Climate and Applied Meteorology

26, 38–52.

Glaccum, R.A., Prospero, J.M., 1980. Saharan aerosols over the

tropical north Atlantic-Mineralogy. Marine Geology 37,

295–321.

Hanel, G., 1988. Single scattering albedo, asymmetry Parameter,

apparent refractive index, and apparent soot content of dry

atmospheric particles. Applied Optics 27, 2287–2294.

Hanel, G., 1994. Optical properties of atmospheric particles:

complete sets obtained through polar photometry and an

improved inversion technique. Applied Optics 33, 7187–7199.

Hermann, M., Stratmann, F., Wilck, M., Wiedensohler, A., 2001.

Sampling characteristics of an aircraft-borne aerosol inlet

system. Journal of Atmospheric and Oceanic Technology 18

(1), 7–19.

Hoornaert, S., Godoi, R.H.M., Grieken, R.V., 2003. Single

particle characterisation of the aerosol in the marine

boundary layer and free troposphere over tenerife, NE

Atlantic, during ACE-2. Journal of Atmospheric Chemistry

46, 271–293.

Horvath, H., 1993. Atmospheric light absorption—a review.

Atmospheric Environment 27A, 293–317.

IPCC, 2001. Radiative Forcing of Climate Change, Climate

Change 2001. The Scientific Basis. Cambridge University

Press, Cambridge, pp. 349–416.

Kalashnikova, O.V., Sokolik, I.N., 2002. Importance of shapes

and compositions of wind-blown dust particles for remote

sensing at solar wavelengths. Geophysical Research Letters 29

(10).

Kallos, G., Kotroni, V., Lagouvardos, K., Papadopoulos, A.,

1998. On the long-range transport of air pollutants from

Europe to Africa. Geophysical Research Letters 25 (5),

619–622.

Levin, Z., Ganor, E., 1996. The effects of desert particles on

cloud and rain formation in the eastern Mediterranean. In:

Guerzoni, S., Chester, R. (Eds.), The Impact of Desert Dust

Across the Mediterranean. Kluwer Academic Publishers,

Dordrecht, pp. 77–86.

Levin, Z., Teller, A., Ganor, E., Yin, Y., 2005. On the

interactions of mineral dust, DEA-salt particles, and clouds:

a measurement and modeling study from the Mediterranean

Israeli Dust Experiment campaign. Journal of Geophysical

Research 110, D20202.

Liao, H., Seinfeld, J.H., 1998. Radiative forcing by mineral dust

aerosols: sensitivity to key variables. Journal of Geophysical

Research 103 (D24), 31637–31645.

Mahowald, N.M., Luo, C., 2003. A less dusty future? Geophy-

sical Research Letters 30 (17), 1903.

Mahowald, N.M., Rivera Rivera, G.D., Luo, C., 2004. Comment

on ‘‘Relative importance of climate and land use in

determining present and future global soil dust emission’’ by

I. Tegen et al. Geophysical Research Letters 31, L24105.

Matsuki, A., et al., 2007. Mixing states and hygroscopicity of

aerosol particles in West Africa: based on AMMA aircraft

campaign in summer 2006. Geophysical Research Abstracts 9,

04729.

ARTICLE IN PRESSK. Kandler et al. / Atmospheric Environment 41 (2007) 8058–80748074

Miehe, G., 1997. Algorithmen mit geringem Informationsbedarf

zur hasenanalyse und Gitterkonstantenbestimmung aus Elek-

tronenbeugungsdaten. Berichte der Deutschen Mineralo-

gischen Gesellschaft, Supplement to European Journal of

Mineralogy 9, 250.

Molinaroli, E., 1996. Mineralogical characterisation of Saharan

dust. In: Guerzoni, S., Chester, R. (Eds.), The Impact of

Desert Dust Across the Mediterranean. Kluwer Academic

Publishers, Dordrecht, pp. 153–162.

Monna, W.A.A., Driedonks, A.G.M., 1979. Experimental data

on the dynamic properties of several propeller vanes. Journal

of Applied Meteorology 18, 699–702.

Nickovic, S., Kallos, G., Papadopoulos, A., Kakaliagou, O.,

2001. A model for prediction of desert dust cycle in the

atmosphere. Journal of Geophysical Research 106 (D16),

18113–18129.

Patterson, E.M., Gillette, D.A., Stockton, B.H., 1977. Complex

index of refraction between 300 and 700 nm for Saharan

aerosols. Journal of Geophysical Research 82 (21),

3153–3160.

Petit, R.H., et al., 2005. Transport of Saharan dust over the

Caribbean Islands: study of an event. Journal of Geophysical

Research 110.

Prospero, J.M., 1999. Long-range transport of mineral dust in the

global atmosphere: impact of African dust on the environ-

ment of the southeastern United States. Proceedings of the

National Academy of Sciences USA 96 (7), 3396–3403.

Prospero, J.M., Ginoux, P., Torres, O., Nicholson, S.E., Gill,

T.E., 2002. Environmental characterization of global sources

of atmospheric soil dust identified with the Nimbus 7 total

ozone mapping spectrometer (TOMS) absorbing aerosol

product. Reviews of Geophysics 40, 1002.

Raabe, O.G., Braaten, D.A., Axelbaum, R.L., Teague, S.V.,

Cahill, T.A., 1988. Calibration studies of the DRUM

impactor. Journal of Aerosol Science 19, 183–195.

Reid, E.A., et al., 2003a. Characterization of African dust

transported to Puerto Rico by individual particle and size

segregated bulk analysis. Journal of Geophysical Research

108 (D19), 8591.

Reid, J.S., et al., 2003b. Comparison of size and morphological

measurements of coarse mode dust particles from Africa.

Journal of Geophysical Research 108 (D19), 8593.

Rodrıguez, S., Torres, C., Guerra, J.C., Cuevas, E., 2004.

Transport pathways of ozone to marine and free-tropospheric

sites in Tenerife, Canary Islands. Atmospheric Environment

38, 4733–4747.

Schutz, L., 1989. Atmospheric mineral dust-properties and source

markers. In: Leinen, M., Sarnthein, M. (Eds.), Paleoclimato-

logy and Paleometeorology: Modern and Past Patterns of

Global Atmospheric Transport. Kluwer Academic Publishers,

Dordrecht, pp. 359–383.

Schutz, L., Sebert, M., 1987. Mineral aerosols and source

Identification. Journal of Aerosol Science 18, 1–10.

Sokolik, I.N., Toon, O.B., 1999. Incorporation of mineralogical

composition into models of the radiative properties of mineral

aerosol from UV to IR wavelengths. Journal of Geophysical

Research 104 (D8), 9423–9444.

Sokolik, I.N., et al., 2001. Introduction to special section:

outstanding problems in quantifying the radiative impacts

of mineral dust. Journal of Geophysical Research 106 (D16),

18015–18027.

Stuut, J.B., et al., 2005. Provenance of present-day eolian dust

collected off NW Africa. Journal of Geophysical Research

110 (D04202).

Tegen, I., Werner, M., Harrison, S.P., Kohfeld, K.E., 2004a.

Relative importance of climate and land use in determining

present and future global soil dust emission. Geophysical

Research Letters 31, L05105.

Tegen, I., Werner, M., Harrison, S.P., Kohfeld, K.E., 2004b.

Reply to comment by N.M. Mahowald et al. on ‘‘Relative

importance of climate and land use in determining present

and future global soil dust emission’’. Geophysical Research

Letters 31, L24106.

Turner, J.R., Hering, S.V., 1987. Greased and oiled substrates as

bounce-free impaction surfaces. Journal of Aerosol Science 18

(2), 215–224.

Usher, C.R., Michel, A.E., Stec, D., Grassian, V.H., 2003.

Laboratory studies of ozone uptake on processed mineral

dust. Atmospheric Environment 37, 5337–5347.

Wernli, H., Davies, H.C., 1997. A Lagrangian-based analysis of

extratropical cyclones. 1. The method and some applications.

Quarterly Journal of the Royal Meteorological Society 123

(538), 467–489.

Wieringa, J., 1967. Evaluation and design of wind vanes. Journal

of Applied Meteorology 6, 1114–1122.

Winkler, P., 1974. Relative humidity and the adhesion of

atmospheric particles to the plates of impactors. Journal of

Aerosol Science 5, 235–240.

Wurzler, S., Reisin, T.G., Levin, Z., 2000. Modification of

mineral dust particles by cloud processing and subsequent

effects on drop size distributions. Journal of Geophysical

Research 105 (D4), 4501–4512.