Tellus (2009), 61B, 51–63
C
2008 The Authors
Journal compilation
C
2008 Blackwell Munksgaard
Printed in Singapore. All rights reserved
TELLUS
State of mixing, shape factor, number size distribution,
and hygroscopic growth of the Saharan anthropogenic
and mineral dust aerosol at Tinfou, Morocco
By N. KAADEN
1∗
, A. MASSLING
1,2
,A. SCHLADITZ
1
,T.M
¨
ULLER
1
,K. KANDLER
3
,
L. SCH
¨
UTZ
4
,B.WEINZIERL
5
, A. PETZOLD
5
,M. TESCHE
1
, S. LEINERT
6
, C. DEUTSCHER
4
,
M. EBERT
3
,S.WEINBRUCH
3
andA. WIEDENSOHLER
1
,
1
Leibniz Institute for Tropospheric
Research, Leipzig, Germany;
2
Aarhus University, National Environmental Research Institute, Department of
Atmospheric Environment, Roskilde, Denmark;
3
Institute for Applied Geosciences, Darmstadt University of
Technology, Darmstadt, Germany;
4
Institute for Atmospheric Physics, Johannes-Gutenberg-University, Mainz,
Germany;
5
Deutsches Zentrum f
¨
ur Luft- und Raumfahrt, Institut f
¨
ur Physik der Atmosph
¨
are, Oberpfaffenhofen, 82234
Wessling, Germany;
6
Environmental Protection Agency, Richview, Dublin, Ireland
(Manuscript received 6 February 2008; in final form 18 August 2008)
ABSTRACT
The Saharan Mineral Dust Experiment (SAMUM) was conducted in May and June 2006 in Tinfou, Morocco.
A H-TDMA system and a H-DMA-APS system were used to obtain hygroscopic properties of mineral dust parti-
cles at 85% RH. Dynamic shape factors of 1.11, 1.19 and 1.25 were determined for the volume equivalent diameters
720, 840 and 960 nm, respectively.
During a dust event, the hydrophobic number fraction of 250 and 350 nm particles increased significantly from 30
and 65% to 53 and 75%, respectively, indicating that mineral dust particles can be as small as 200 nm in diameter. Log-
normal functions for mineral dust number size distributions were obtained from total particle number size distributions
and fractions of hydrophobic particles. The geometric mean diameter for Saharan dust particles was 715 nm during the
dust event and 570 nm for the Saharan background aerosol.
Measurements of hygroscopic growth showed that the Saharan aerosol consists of an anthropogenic fraction (pre-
dominantly non natural sulphate and carbonaceous particles) and of mineral dust particles. Hygroscopic growth and
hysteresis curve measurements of the ‘more’ hygroscopic particle fraction indicated ammonium sulphate as a main com-
ponent of the anthropogenic aerosol. Particles larger than 720 nm in diameter were completely hydrophobic meaning
that mineral dust particles are not hygroscopic.
1. Introduction
The Saharan desert is globally one of the major natural sources
for mineral dust particles. The source strength varies between
130 and 5000 Tg
.
yr
−1
(Swap et al., 1996; Cakmur et al., 2006;
Goudi and Middleton, 2006). Dust plumes spreading from North
Africa over the Atlantic Ocean are observed during the whole
year (Moulin et al., 1997; Engelstaedter et al., 2006).
Maring et al. (2003) investigated particle number size distri-
butions of Saharan dust in the outflow from the African continent
at the Canary Islands and Puerto Rico. They found invariant par-
ticle number size distributions for both locations until a size of
∗
Corresponding author.
e-mail: kaaden@tropos.de
DOI: 10.1111/j.1600-0889.2008.00388.x
7.3 μm. Only for larger diameters (>7.3 μm), a change in the
distribution was observed as particles were removed most prob-
ably by sedimentation from the atmosphere during the transport
over the Atlantic Ocean. However, up to now there is no data
on particle number size distributions of Saharan aerosol in the
source region obtained within the last 10 yr.
Particle number size distributions are crucial for determining
aerosol radiative impacts (Tegen and Fung, 1994). Optical prop-
erties depend on the size and shape of the particles (Covert et al.,
1972; Quinn et al., 1996; Tegen et al. 1996; Heintzenberg et al.,
1997), which in turn vary due to hygroscopic growth. Especially
for the understanding of remote sensing data, the knowledge
of particle shape is of great interest in terms of their optical
properties. Dubovic et al. (2002) showed that in comparison
to Mie theory, a model simulating the irregularity of a particle
improves the dust-particle phase function and size distribution.
Tellus 61B (2009), 1 51
PUBLISHED BY THE INTERNATIONAL METEOROLOGICAL INSTITUTE IN STOCKHOLM
SERIES B
CHEMICAL
AND PHYSICAL
METEOROLOGY
52 N. KAADEN ET AL.
Mishchenko et al. (1997) also reported that the phase function is
influenced by the particle shape and concluded that Mie theory
cannot be used to calculate the phase function of dust particles.
Dust particles provide reactive sites for heterogeneous reac-
tions with, for example, HNO
3
(Bauer et al., 2004). Acidification
of mineral dust particles, for example, by nitric acid may lead
to a significant uptake of water (Laskin et al., 2005) and thus to
an increase in size and a change in shape. Levin et al. (1996)
stated that desert dust particles larger than 1 μm in diameter are
coated with sulphate and other soluble materials after contact
with clouds (cloud processing). Since the particles were trans-
ported hundreds of kilometres, they are liable to lots of aging
processes. There are speculations if dust particles in the source
region already consist of a certain mass fraction of soluble ma-
terial. However, there are no measurements about hygroscopic
growth for Saharan dust particles in the source region. Hygro-
scopic properties of the Asian desert dust particles have been
investigated in the outflow of China (Massling et al., 2007).
No hygroscopic growth was observed there for the Asian dust
particles.
Knowledge on the state of mixing of aerosol particles provides
information on sources, transformation, and aging processes of
the aerosol population. An important issue concerning the state
of mixing was shown by Wang and Martin (2007), who dis-
cussed aerosol optical properties via satellite observations. They
detected that the change from internal to external mixing for
the insoluble and soluble materials changed the single scattering
albedo, the aerosol optical depth, and the aerosol effective radius.
For the interpretation of optical properties of ambient aerosols
from satellite measurements, it is thus of great importance to
know how the aerosol is mixed. Wang and Martin (2007) also
comment that assumptions about the state of mixing in retrieval
algorithms affect the deviated number concentration.
A comprehensive field study close to the Saharan desert was
conducted to investigate and better understand the shape factor,
number size distributions, state of mixing, and the hygroscopic
growth of the Saharan aerosol for background and dusty condi-
tions. During May and June, 2006, in situ ground-based mea-
surements were performed as part of the Saharan Mineral Dust
Experiment (SAMUM). Particle number size distributions have
been measured directly in the size range from 20 nm up to 10 μm.
Shape factors, state of mixing, size distributions of hydrophobic
particles and hygroscopic growth of individual particle groups
were derived from size-resolved hygroscopicity measurements
inthesizerangefrom30nmupto1μm in dry diameter.
2. Methods
2.1. Location
For the SAMUM field experiment, a location close to the Saharan
desert was chosen, which also provided a certain infrastructure
for the measurement container. The ‘Kasbah Hotel Porte au
0
45
90
135
180
225
270
315
1
2
3
4
5
16.5 km
Fig. 1. Wind rose for the whole measuring period with the hotel (3)
being situated in the center of the chart. The black line describes the
road from Quarzazate (1) over Zagora (2) passing Tinfou (4) and ends
in M’Hamid (5).
Sahara’, a hotel near the small village Tinfou (30.14
◦
N, 5.36
◦
W;
684 m a.s.l.) 35 km southeast of Zagora in Southern Morocco
was thus chosen as measurement site. The Saharan desert begins
approximately 60 km south of Zagora.
Two main wind directions prevailed during the field experi-
ment which is shown as wind rose in Fig. 1. Next to the wind
rose, the street from Quarzazate (1) with adjacent villages and
the hotel (3) in the centre of the chart are shown. The container
was located about 100 m northeast of the hotel. Since the ho-
tel used mostly electrical power, contamination during the mea-
surements occurred most probably only from the street. Data did
undergo a quality check (e.g. peaks in number concentration), so
that strong local contamination is excluded for the results shown
here. A more detailed description of the measuring site is given
in Heintzenberg (2008) and Kandler et al. (2008).
2.2. Instrumentation
All instruments were placed inside an air-conditioned sea con-
tainer. To avoid strong heating inside the container, white
awnings were placed around the container to reflect incoming
solar radiation. An inlet system consisting of a low flow PM10
impactor was used to collect the ambient aerosol. Inside of the
container, the aerosol was conducted to the individual instru-
ments.
Two systems were applied, a Hygroscopicity-Tandem Differ-
ential Mobility Analyzer (H-TDMA; Massling et al., 2003) and
a Hygroscopicity-Differential Mobility Analyzer-Aerodynamic
Tellus 61B (2009), 1
STATE OF MIXING, SHAPE FACTOR, NUMBER SIZE DISTRIBUTION, AND HYGROSCOPIC GROWTH 53
ia htaehSr
AMDT-H
ia htaehSr
AMD
emidmu
CPC
yrd
CPC
hu fidimide
ria ssecxE
i
lo
s
o
re
An
ibma HR()tne
co tidngnin
oi
lartueNirez
)yrd HR(
co tidngninoi
A
M
D
emidmu
)denifed HR(
ria ssecxE
(e
d
H
Rfi)den
i
a
ht
a
eh
S
r
FAH
A
M
D
SPA
yrd
SP
A
fidimuhide
)etatsretni HR(
ria ss
ecxE
i losoreAn
(ed HRfiden))tneibma HR(
no
i
t
a
cif
i
dimu
h
itid
noci
nogn
lartueNirez
)yrd HR(
ia htaehSr
itidnocinogn
(fed HR
i
)den
SPA-AMD-H
Fig. 2. Schematic set-up of the
Hygroscopicity-Tandem Differential
Mobility Analyzer (H-TDMA) and the
Hygroscopicity-Differential Mobility
Analyzer-Aerodynamic Particle Sizer
(H-DMA-APS), the last one including a
custom-made high aerosol flow-DMA
(HAF-DMA).
Particle Sizer (H-DMA-APS; Leinert and Wiedensohler, 2008)
to determine the state of mixing, shape factor, size distributions
of hydrophobic particles, and hygroscopic growth of individual
particle groups. These two systems work similarly with dif-
ferences in the sampling technique and operating size range.
Figure 2 illustrates a schematic set-up for both instruments.
The aerosol is dried by passing an aerosol diffusion dryer first
and then charged in a custom-made Kr85 neutralizer to bring
the particles into the bipolar charge equilibrium.
2.2.1. H-TDMA Quasi-monodisperse (certain electrical mo-
bility) particle fractions (mobility diameters of 30, 50, 80, 150,
250 and 350 nm) are selected in the first Differential Mobility
Analyzer (DMA; Knutson and Whitby, 1975; here: type Hauke
medium). These quasi-monodisperse particles are divided into
two parts afterwards. One part is conducted to a Condensa-
tion Particle Counter (CPC, TSI 3010, TSI Inc., St. Paul, MN,
USA) that measures the number concentration of the selected
dry particles. The second part is conditioned to a defined rela-
tive humidity (RH) by a humidity conditioner based on a Nafion
membrane, in this case to 85%. The second DMA (again Type
Hauke medium) together with the second CPC (again TSI 3010)
is used as a mobility size spectrometer measuring the number
size distribution of the humidified aerosol at 85% RH.
2.2.2. H-DMA-APS The H-DMA-APS is similar in function,
but there are slight differences. The particles are selected in the
mobility diameter range around 1 μm using a custom-made high
aerosol flow—DMA (HAF-DMA) designed for an aerosol flow
rate of 2 l min
−1
andsheathairflowrateof20lmin
−1
(Leinert
and Wiedensohler, 2008). Again, the quasi-monodisperse
aerosol is divided into two parts downstream of the DMA. The
first fraction of the aerosol is led to an Aerodynamic Particle
Sizer (APS, TSI 3320, TSI Inc., St. Paul MN, USA) to de-
termine the dry number size distribution (aerodynamic diame-
ter) of the selected dry mobility diameters (here 800, 1000 and
1200 nm). The second aerosol fraction is conditioned by Nafion
membranes to 85% RH. The number size distribution of the hu-
midified aerosol is then obtained in a second APS operated with
a sheath air flow humidified to 85%, too.
For both systems, regular calibration scans with ammonium
sulphate were performed to recalculate the real humidity inside
the second DMA (H-TDMA) and the second APS (H-DMA-
APS). Furthermore, polystyrene latex (PSL) scans in the
H-DMA-APS were taken to correct for size shifts of the APS
instruments.
Next to the instruments for measuring the hygroscopicity,
also a differential mobility particle sizer (DMPS; Birmili et al.,
1999) in combination with an APS was applied. With these in-
struments, the particle number size distribution in the range from
20 nm up to 10 μm was obtained. Detailed information about
the particle number size distribution can be found in Schladitz
et al. (2008).
2.3. Calculations
Atmospheric aerosol particles are generally divided into groups
of different hygroscopic growth depending on their origin and
transformation history. From the H-TDMA and H-DMA-APS
data, hygroscopic growth factors (gf) of different hygroscopic
particle groups and their corresponding number fractions (nf)
can be determined. Additionally, the dynamic shape factor (χ )
can be determined from the H-DMA-APS measurements.
The hygroscopic growth factor is defined as the ratio of the
humidified (wet) to the dry diameter of the same particle:
gf =
d
ve,wet
d
ve,dry
, (1)
with d
ve
being the volume equivalent diameter. The individ-
ual systems use different diameter definitions for sizing. DMA
Tellus 61B (2009), 1
54 N. KAADEN ET AL.
measurements are based on the mobility diameter (d
m
), which
can easily be converted into a volume equivalent diameter by
knowledge of the particle shape factor (Hinds, 1999; DeCarlo
et al., 2004). The APS measures an aerodynamic diameter (d
a
).
To calculate the volume equivalent diameter from d
a
, the particle
density (ρ
P
) and the shape factor must be known. Both equa-
tions are described below, neglecting the Cunningham correction
since it changes results in the third decimal place.
d
ve
=
d
m
χ
, (2)
d
ve
= d
a
χ.
ρ
0
ρ
P
. (3)
Hereby, ρ
0
is the unit density of 1.0 g cm
−3
. The density of
dry particles is obtained as material density from the electron-
microscopic single particle analysis (Kandler et al., 2008). This
material density and the effective density for aerodynamic pro-
cesses are not necessarily equal, for example for agglomerates.
Since agglomerates which influence the particle density are not
the majority found in the samples it is stated, that the mineral
density is realistic. Combining eqs (2) and (3) results in an equa-
tion for the dynamic shape factor:
χ =
3
d
2
m
d
2
a
.
ρ
P
ρ
0
. (4)
Since d
m
is derived for dry particles, χ obtained by eq. (4) is
only valid for dry particles.
With the H-TDMA, particles in the size range smaller than
350 nm are measured. Because the particles are relatively small,
the shape of these particles can be assumed to be spherical. This
assumption was confirmed by a study of Wu and Colbeck (1996).
There, a linear relationship between the logarithm of d
ve
and the
logarithm of χ was found. This calculation was applied to the re-
sults obtained by the H-DMA-APS measurements showing that
the dynamic shape factor reaches Unity at approximately 500
nm in volume equivalent diameter (see Section 3.1). Therefore,
the mobility diameter for particles smaller than 500 nm equals
the volume equivalent diameter and the growth factor can be
calculated as shown in eq. (1).
The situation for the H-DMA-APS is more complicated. Cal-
culation of the growth factor also follows eq. (1), but d
ve
first
has to be substituted with d
a
for every single scan, leading
to eq. (5):
gf =
d
ve,wet
d
ve,dry
=
d
a,wet
d
a,dry
.
χ
wet
.ρ
P,dry
χ
dry
.ρ
P,wet
= gf
a
.
χ
wet
.ρ
P,dry
χ
dry
.ρ
P,wet
,
(5)
with gf
a
being the aerodynamic growth factor.
Equation (5) includes two unknown parameters, the wet den-
sity (ρ
P,wet
) and the wet shape factor (χ
wet
). ρ
P,dry
and χ
dry
are
obtained as mentioned above. In the case that the particles do not
grow by water uptake, ρ
P,wet
and χ
wet
equal ρ
P,dry
and χ
dry
,re-
spectively. The growth factor is equal to the aerodynamic growth
factor. In the case that the particles are strongly growing, χ
wet
is Unity and ρ
P,wet
has to be assigned iteratively and converges
to that of pure water at high RH. The iteration is done by the
assumption that ρ
P,wet
is the ratio of the added masses (of the
dry particle and the condensed material, which is in this case
water) and the added volumes (of the dry particle and the con-
densed material). This assumption is included in eq. (5) resulting
in eq. (6). For all other cases, both wet parameters have to be
determined iteratively. Therefore, χ
wet
has to be assigned under
reasonable physical considerations. Here, a simple parameteri-
zation of χ
wet
is used following the condition that in the growth
factor range between 1.1 and 1.4, χ
wet
converges from χ
dry
to
Unity in hyperbola form as expressed in eq. (7).
gf
a
=
χ
wet
χ
dry
·
1 + (gf
3
− 1)
ρ
0
ρ
dry
gf
, (6)
χ
wet
=
χ
dry
− 1
10
+ 1 (for1.1 < gf
a
< 1.4). (7)
Usually, more than one particle group in terms of hygroscopic
growth is observed in particle number size distributions at high
RH. Particles are usually separated into a ‘more’ hygroscopic,
a ‘less’ hygroscopic and a hydrophobic particle group. Classifi-
cation was done based on measured hygroscopic growth factors
as follows: hydrophobic (gf < 1.15), ‘less’ hygroscopic (1.15 <
gf < 1.45), and ‘more’ hygroscopic (gf > 1.4) particle fractions.
The integral of the humidified particle number size distribution
delivers the total number of particles N observed during the
measurement. Thus, the particle number fraction of individual
hygroscopic particle groups nf
i
can be obtained by the ratio of
number of particles of this individual group N
i
and the total
number of particles N
tot
, following eq. (8):
nf
i
=
N
i
N
tot
. (8)
Uncertainties of the measured and calculated parameters were
obtained by an error estimate and by propagation of errors,
respectively. This leads to an error of approximately 4% for
the hygroscopic growth factor for the H-TDMA after correction
for the DMA shift and 14% for the number fraction. For the
H-DMA-APS, the error for the growth factor is approximately
1.5%. There is an error of about 10% for the number fraction,
although just one mode was found leading to a number fraction
of 1. The error for the shape factor is approximately 15%, taking
the uncertainty of the dry density of 15% into account.
2.4. Instrumentation and methods of project partners
Next to the results from hygroscopic measurements, results
from other measurement techniques are compared. On board
the ‘Falcon’ research aircraft of the DLR (Deutsches Zentrum
f
¨
ur Luft- und Raumfahrt), extensive aerosol in situ instru-
mentation was operated (Weinzierl et al., 2008). Volatility
Tellus 61B (2009), 1
STATE OF MIXING, SHAPE FACTOR, NUMBER SIZE DISTRIBUTION, AND HYGROSCOPIC GROWTH 55
measurements were performed in the diameter range from 10
nm to 2.5 μm by a combination of particle sizing instruments
with a thermal denuder (TD) (Clarke, 1991). The heating tem-
perature of the TD was set to 250
◦
C in order to remove high to
medium volatility organics, sulphuric acid and ammonium sul-
phate while leaving behind compounds with lower vapourization
pressures like BC, sea salt, dust and crust material (Pinnick et al.,
1987; Rose et al., 2006). More detail on the airborne volatility
measurements performed during SAMUM is given in Weinzierl
et al. (2008).
Other measurements were also performed at the airport of
Quarzazate. The main focus at this location was remote sens-
ing. One of the three lidars at Quarzazate airport was the
IfT-six-wavelength Raman lidar Backscatter Extinction lidar-
Ratio Temperature Humidity profiling apparatus (BERTHA;
Althausen et al., 2000). With this lidar, laser pulses at 355,
400, 532, 710, 800 and 1064 nm wavelength are transmitted.
The Raman lidar BERTHA was used to obtain profiles of the
backscatter coefficient, the extinction coefficient, the extinction-
to-backscatter ratio (lidar-ratio), and the depolarization ratio
(Tesche et al., 2008).
At the hotel site near Tinfou, particle sampling was per-
formed parallel to measurements with the H-TDMA and the
H-DMA-APS. Analysis of these samples was amongst others
conducted with electron-microscopic single particle analysis.
With scanning electron microscopy (FEI ESEM 310 Quanta
200 FEG, Eindhoven, The Netherlands), several thousand par-
ticles were analyzed. The size of the particles is given in vol-
ume equivalent diameter (Kandler et al., 2008). The morphol-
ogy and chemical composition of the particles was determined
by energy-dispersive X-ray microanalysis. For smaller particles
(<700 nm), this had to be done manually, since automatic anal-
ysis is not reliable.
2.5. Air mass classification
During the four week field study in May–June 2006, several
changes in air mass derivation occurred. Different origins of the
138 140 142 144 146 148 150 152 154 156 158
0.0
0.2
0.4
day of year
0
200
400
600
0
1000
2000
0
180
360
0
5
10
15
v / [m/s]
dd / [°]
GC
dp < 0.1 μm
dp > 0.6 μm
Fig. 3. Time series of wind speed (v) and
wind direction (dd), number concentrations
(N) of particles smaller than 100 nm and
larger than 600 nm and soot mass
concentrations (m
GC
). The shaded areas
present the different periods for the Saharan
background (areas 1 and 3) and the dust
event (area 2).
Saharan aerosol were identified based on meteorological time
series and 2-d backward trajectories (Knippertz et al., 2008).
Although some errors can occur regarding backward trajectories,
the time frame of 2 d is small enough to get useful information
about aerosol source and transport. Two representative periods
have been selected for the whole time period to characterize
different aerosol types during the experiment.
(1) Saharan background: This first classification describes
the main streaming observed during the SAMUM field study.
Two separated time periods, 16 to 19 May and 2 to 7 June
2006 were chosen to describe this air mass type. The air passed
over Algeria and arrived in Tinfou from eastward direction.
Sometimes, these air masses originated from the northeastern
part of North Africa and crossed the Mediterranean Sea during
advection to the measuring place. The wind velocity, recorded
by an acoustic anemometer, reached a value of approximately
5ms
−1
and the visibility amounted to 10–30 km in average
observed by a VPF-710 visibility sensor.
(2) Mineral dust event: One strong local dust event occurred
from 22 to 26 May 2006. Backward trajectories did not show
a distinct origin, but some trajectories arrived from southern
directions and therewith from more central parts of the Saharan
desert. The wind velocity reached a value of 16 m s
−1
and the
mean visibility was below 10 km.
3 Results and discussion
As mentioned above, all particle diameters presented here are
converted from mobility or aerodynamic diameter to volume
equivalent diameter. The reason for this conversion is that optical
properties depend on the particle volume rather than on their
mobility or aerodynamic properties.
To get a picture of the whole campaign data at a glance,
wind speed and direction as well as number concentration for
particles smaller than 100 nm and larger than 600 nm and the
soot concentration are plotted in Fig. 3. In this figure, also the
two representative periods are shown.
Tellus 61B (2009), 1
