Comparison of optical and microphysical properties of pure
Saharan mineral dust observed with AERONET Sun photometer,
Raman lidar, and in situ instruments during SAMUM 2006
D. Müller,1,2K.-H. Lee,1J. Gasteiger,3M. Tesche,4B. Weinzierl,5,6K. Kandler,7
T. Müller,2C. Toledano,8,9S. Otto,10,11D. Althausen,2and A. Ansmann2
Received 2 September 2011; revised 13 January 2012; accepted 15 January 2012; published 14 April 2012.
 The Saharan Mineral Dust Experiment (SAMUM) 2006, Morocco, aimed at the
characterization of optical, physical, and radiative properties of Saharan dust. AERONET
Sun photometer, several lidars (Raman and high-spectral-resolution instruments), and
airborne and ground-based in situ instruments provided us with a comprehensive set of
data on particle-shape dependent and particle-shape independent dust properties. We
compare 4 measurement days in detail, and we carry out a statistical analysis for some
of the inferred data products for the complete measurement period. Particle size
distributions and complex refractive indices inferred from the Sun photometer observations
and measured in situ aboard a research aircraft show systematic differences. We find
differences in the wavelength-dependence of single-scattering albedo, compared to
light-scattering computations that use data from SOAP (spectral optical absorption
photometer). AERONET data products of particle size distribution, complex refractive
index, and axis ratios were used to compute particle extinction-to-backscatter (lidar) ratios
and linear particle depolarization ratios. We find differences for these parameters to
lidar measurements of lidar ratio and particle depolarization ratio. Differences particularly
exist at 355 nm, which may be the result of differences of the wavelength-dependent
complex refractive index that is inferred by the methods employed in this field campaign.
We discuss various error sources that may lead to the observed differences.
Citation: Müller, D., et al. (2012), Comparison of optical and microphysical properties of pure Saharan mineral dust observed
with AERONET Sun photometer, Raman lidar, and in situ instruments during SAMUM 2006, J. Geophys. Res., 117, D07211,
 This contribution is part three of our study [Müller
et al., 2010a, 2010b] of optical and microphysical proper-
ties of mineral dust observed during the Saharan Mineral
Dust Experiment (SAMUM) 2006. The field experiment
was carried out in Morocco.
 The main results of SAMUM 2006 are presented in a
collection of papers in a special issue in Tellus, 61B, 2009.
We operated three Raman aerosol lidars and two Sun pho-
tometers at Ouarzazate airport (30.93° N, 6.9° W, 1133 m
above sea level (asl)) from 11 May until 10 June 2006
[Tesche et al., 2009; Toledano et al., 2009]. One of the Sun
photometers was an Aerosol Robotic Network (AERONET)
system [Holben et al., 1998]. The second instrument was
operated by the University of Munich.
 Optical and microphysical properties of Saharan min-
eral dust were inferred from observations of direct and dif-
fuse solar radiation and the use of AERONET’s latest
mineral dust model [Dubovik et al., 2006]. This model has
been specifically designed for the analysis of particles of
non-spherical shape. Spheroidal particle shape is assumed in
the retrieval procedure.
 Airborne measurements were carried out with the
Falcon aircraft between 18 May and 7 June 2006 [Petzold
et al., 2009; Weinzierl et al., 2009]. The aircraft carried a
high-spectral-resolution lidar [Esselborn et al., 2009] which
allowed us to sound the atmosphere in close vicinity around
1Atmospheric Remote Sensing Laboratory, School of Environmental
Science and Engineering, Gwangju Institute of Science and Technology,
Gwangju, South Korea.
2Leibniz Institute for Tropospheric Research, Leipzig, Germany.
4Department of Applied Environmental Science, Stockholm University,
5Institut für Physik der Atmosphäre, Deutsches Zentrum für Luft- und
Raumfahrt, Wessling, Germany.
Universität Darmstadt, Darmstadt, Germany.
9Formerly at Meteorological Institute, Ludwig Maximilian University,
11Formerly at Institut für Methodik der Fernerkundung, Deutsches
Zentrum für Luft- und Raumfahrt, Wessling, Germany.
Copyright 2012 by the American Geophysical Union.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117, D07211, doi:10.1029/2011JD016825, 2012
1 of 25
the ground station. Aircraft measurements provide us with
particle size distributions in the size range up to 30 mm
radius. Chemical analysis of thousands of particles collected
aboard the aircraft allowed us to compute the complex
refractive index and the axis ratio distribution of the particles
[Kandler et al., 2009]. We measured with lidar the particle
extinction coefficients at 355 and 532 nm, and thus the
extinction-related Ångström exponent for this wavelength
pair. Both parameters are less influenced by particle shape.
We also measured particle backscatter coefficients, lidar
ratios, and linear particle depolarization ratios, which sig-
nificantly depend on particle shape.
 To date we lack in high-quality experimental field data
that allow us to test the robustness of light-scattering models
of mineral dust. Müller et al. [2010a] present a comparison
of optical parameters of mineral dust for which particle
shape may not be so important in the measurement process,
respectively, the retrieval procedure. Müller et al. [2010a]
and Müller et al. [2010b] present results of one measure-
ment case only. We selected 19 May 2006. It was the day for
which we had data from all measurement platforms.
 We compared data products acquired with Raman and
high-spectral resolution lidar, Sun photometer and airborne
and ground-based in situ instrumentation. One purpose of
our study was to test the quality of the derived data products.
Measurement and data analysis techniques employed in this
campaign are affected by particle shape. Thus, differences of
the same data product inferred by different measurement
platforms may help in identifying if particle shape or other
reasons may be responsible for the observed deviations.
 One striking result was that particle size distribution
inferred from AERONET Sun photometer measurements
and measured in situ aboard a research aircraft differed
significantly [Müller et al., 2010a]. We found that so-called
shape-independent parameters, i.e. dust properties whose
determination with the various instruments platforms
depends little on the exact knowledge of particle shape agree
reasonably well [Müller et al., 2010a]. These parameters
are particle extinction, scattering and absorption coefficients,
the Ångström exponents that are associated with the three
latter parameters, and single-scattering albedo. Particle
shape-dependent parameters in contrast showed in part
large differences, which may be caused by an insufficient
knowledge of particle shape and/or the light-scattering
models that were used in data analysis [Müller et al., 2010b].
These parameters are particle extinction-to-backscatter
(lidar) ratio, linear particle depolarization ratios, and aspect
ratios of dust; in this latter case we approximate the complex
shape of dust particles by spheroids.
 Drawback of our previous study is the fact that we
used only one measurement day for the comparison study.
At the time of our data analysis we did not have the set of
parameters available for further studies. In the past 1.5 years
we analyzed the data of the SAMUM 2006 campaign in a
rather comprehensive fashion. We collected as many data as
possible of the different measurement platforms and we
carried out the comparison study for many more measure-
ment days. A goal of our study is to find out if the results of
Müller et al. [2010a, 2010b] can be generalized.
 In this contribution we present results of three more
measurement cases in detail. We follow the procedures
outlined by Müller et al. [2010a, 2010b]. We also provide a
statistical analysis of the SAMUM 2006 results. We ana-
lyzed all measurement days of the different instrument
platforms that can be used for our study. Table 1 summarizes
the parameters that we compare.
 We also carried out light-scattering computations as
an extension of the studies presented by Wiegner et al.
. A goal of this work is to see if any combination of
measured dust parameters allows us to reproduce some dust
optical parameters, i.e., linear dust depolarization ratio and
lidar ratio, which can be measured directly with lidar and
without critical assumption on particle shape.
 Our study also expands on the publications by Reid
et al. [2003, 2008]. The authors critically discuss results
of measurements of mineral dust observed with AERONET
Sun photometer and various in situ instrumentations.
Observations were carried out in the Arabian Peninsula
[Reid et al., 2008], which is in the source regions of mineral
dust emissions, and in Puerto Rico [Reid et al., 2003], which
is in the far-field of Saharan dust. The authors find con-
siderable variation of dust size distributions. The authors
critically discuss reasons for the observed variations. Some
of the variations are likely due to the measurement
Table 1. Data Products Compared in Our Studya
Raman Lidar GroundHSRL AircraftAERONET GroundIn Situ AircraftIn Situ Ground
Extinction-to-backscatter (lidar) ratio
Linear particle depolarization ratio
Particle size distribution
Complex refractive index (real part)
Complex refractive index (imaginary part)
Particle aspect ratio
aAll parameters except for the particle size distribution and the aspect ratio are presented as wavelength-dependent quantities. The wavelengths for which
we carried out the analysis are given in the figure captions of this contribution. Instruments and data analysis are described by Tesche et al. ,
Esselborn et al. , Freudenthaler et al. , Heese et al. , Müller et al. [2009, 2010a, 2010b], Weinzierl et al. , and Kandler et al.
. Regarding the Raman lidar measurements, we calculated the Ångström exponents from the profiles of the extinction coefficients measured with
Raman lidar. Regarding the AERONET Sun photometer and the in situ aircraft measurements, we computed the Ångström exponents from particle
extinction, scattering, and absorption coefficients. Details are given in section 3.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
2 of 25
technique that is employed for the measurement of the par-
ticle size distributions. In other words, different measure-
ment techniques may result in different results for the same
particle size distributions.
 Another goal of our comparison study is to develop a
particle light-scattering model for mineral dust. We want to
apply it to the inversion of dust optical properties (measured
with lidar) into dust microphysical properties, e.g., particle
effective radius, volume, surface-area, and number concen-
tration, and the complex refractive index. Until now we can
infer these parameters from multiwavelength Raman lidar
observations only if the particles are of spherical shape
[Müller et al., 1999a; Veselovskii et al., 2002; Böckmann
et al., 2005].
 The light-scattering model that is used in the
AERONET inversion algorithm is one potential candidate
for our inversion method. However, this dust model has not
been designed for describing light-scattering properties at
180°. Some of the data products presented here are not a
standard output of the AERONET retrieval scheme, but they
can be comparably easily calculated from particle size dis-
tribution, complex refractive index, and aspect ratio
[Dubovik et al., 2006]. Just recently this light-scattering
model has, for the first time, been tested in the inversion of
multiwavelength lidar data [Veselovskii et al., 2010].
 We are currently carrying out studies with the inver-
sion methodology of Veselovskii et al. . We use the
concept of the AERONET light-scattering model. Our test
data set consists of several measurement cases of SAMUM
2006 and data taken during the follow up campaign
SAMUM 2008 [Ansmann et al., 2011]. The interpretation of
the inversion results will also rest upon the findings of our
study made here and the previous two contributions by
Müller et al. [2010a, 2010b].
 In section 2 we summarize the instruments and the
methods. In section 3 we present results of several measure-
ment days in more detail. In section 4 we present a statistical
overview on the results of SAMUM 2006. We summarize our
results in section 5. We close in section 6 with an outlook on
new methodologies that could further improve our under-
standing of mineral dust properties.
 We briefly summarize the measurement techniques.
Instruments, data analysis methods, and error estimates are
described in detail in the individual papers of the first special
issue on SAMUM (Tellus, 61B, 2009), and by Müller et al.
 We carried out collocated observations with an
AERONET Sun photometer [Holben et al., 1998, 2001] and
three Raman lidars at Ouarzazate airport. One lidar instru-
ment was the Backscatter Extinction lidar-Ratio Tempera-
ture Humidity profiling Apparatus (BERTHA) of the
Leibniz Institute for Tropospheric Research (IfT) [Althausen
et al., 2000; Tesche et al., 2009]. Two more systems, the
three-wavelength Multiwavelength Lidar System (MULIS)
[Freudenthaler et al., 2009] and the one-wavelength Por-
table Lidar System (POLIS) [Freudenthaler et al., 2009]
were operated by the University of Munich. A fourth lidar
system was operated aboard the DLR (German Aerospace
Center) research aircraft Falcon. This lidar is a high-spectral-
resolution lidar (HSRL) [Esselborn et al., 2009]. The aircraft
passed over the Ouarzazate ground station on several days
during SAMUM. Müller et al. [2010a, Table 2] summarize
the data products that were measured with the lidar systems.
Details of data analysis and error analysis are described by
Tesche et al. , Freudenthaler et al. , Esselborn
et al. , and Heese et al. .
 The AERONET instrument measured the direct radi-
ation of the Sun at 339, 379, 441, 501, 675, 869, 940, 1021,
and 1638 nm wavelength. Sky radiance (almucantar) mea-
surements were carried out at 441, 501, 675, 869, 1021, and
1638 nm. The instrument was calibrated at NASA Goddard
Space Flight Center (GSFC) before and after the field cam-
paign. The accuracy of the aerosol optical depth measure-
ments is approximately 0.01 at visible and near infrared
wavelengths and approximately 0.02 in the ultraviolet
wavelength region [Eck et al., 1999].
 Aerosol optical depth, aerosol phase function, particle
volume size distribution from 0.05–15 mm in radius, and
the complex refractive index in the range from 1.33–1.6 (real
part) and from 0.0005i–0.5i (imaginary part) were derived
from the data. The algorithm is described by Dubovik and
King  and Dubovik et al. . From the retrieved
data products we calculated all other parameters we use
in our study, i.e., particle backscatter coefficients, lidar
ratios, and linear particle depolarization ratios [Müller et al.,
2010b]. The instrument was equipped with a novel mea-
surement channel at 1638 nm. We expect this channel to
improve the retrieval accuracy of some of the data products,
as for instance the particle size distribution.
 Another important output parameterof ourstudyisthe
aspect ratio. Particles of spheroidal shape are assumed in the
analysis of theAERONET data. The aspectratio isdefined as
the ratio of the longest axis of a spheroid to its shortest axis;
notethatthese twoaxesneedtobechosensuchthat bothaxes
are perpendicular to each other. Thus, the aspect ratio is
always ≥1. In contrast, the axis ratio is defined by Dubovik
et al.  as the ratio between the length of the rotational
axis to the length of the axis perpendicular to it. The axis
ratio is <1 for flat spheroids, and >1 for elongated spheroids.
 We point out to some additional error sources. The
data were analyzed by the AERONET team and provided to
us directly. The results may therefore differ from the level 2
results in the AERONET database. Reason for this approach
was that we had the opportunity to incorporate the novel
measurement channel at 1638 nm. The data taken with the
1638-nm channel were added in the signal processing chain.
The data at 339 and 379 nm were not provided to us.
 The accuracy of the calibration of this channel at
1638 nm is generally worse than for the other channels in the
visible spectrum. A comparison of data products inferred
with and without the new channel shows that additional
errors are comparably low in view of all other error sources.
We refer also to our previous publications in which we
present retrieval results with and without this novel mea-
surement channel [Müller et al., 2010a, 2010b]. We
emphasize that this channel still is not part of the standard
data analysis chain of level 2 data, and therefore the results
presented here may differ from a future level 2 data analysis.
 Another source of error in our data analysis may be
low optical depth which prevailed on several of the mea-
surement days such as 19 May and 4 June 2006. In order to
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
3 of 25
qualify for level 2 data analysis, optical depth must be
greater than 0.4 at 440 nm and the solar zenith angle must be
larger than 50° during the measurement. It is clear that this
constraint of high optical depth was not met in many cases
during the measurement period [see, e.g., Müller et al.,
2010a, Figure 1]. In nearly all of the analyzed cases optical
depth was larger than 0.2 at 440 nm. Thus it was also nec-
essary to have the analysis done by the AERONET team
directly, as it may reduce some of the problems in the quality
of the retrieval results.
 The research aircraft Falcon carried the nadir-looking
HSRL and instruments for observations of aerosol particle
microphysical and chemical properties. The instruments are
described by Petzold et al. , Weinzierl et al. ,
and Kandler et al. . We operated another ground-
based station at Tinfou (30.2° N, 5.6° W) for collection and
chemical analysis of the particles from which we inferred the
complex refractive index [Kandler et al., 2009] and single-
scattering albedo [Müller et al., 2009].
 We used scanning electron microscopy to determine
the aspect ratios of the collected particles as a function of
particle size and flight level of the aircraft [Kandler et al.,
2009]. The aspect ratio was determined from an ellipse that
was fitted into the two-dimensional particle outline mea-
sured by the electron microscope. We rotated the ellipse
around its longest half-axis, thus, the axis ratio is always ≥1.
and 3 and 4 June 2006
 Table 2 lists the measurement times of the different
instruments on 19 and 28 May and 3/4 June 2006. Knippertz
et al.  present an overview on the meteorological sit-
uation. Details on the measurement situation, data acquisi-
tion and analysis are found in the publications by Tesche
et al.  (BERTHA), Esselborn et al.  (HSRL),
Freudenthaler et al.  (MULIS), Heese et al. 
(POLIS), Müller et al. [2010a, 2010b] (AERONET),
Weinzierl et al.  (airborne particle size distribution
measurements), Kandler et al.  (chemical analysis of
particles collected aboard the Falcon aircraft and at ground at
Tinfou), and Müller et al.  (ground-based measure-
ments of particle absorption with SOAP).
 The three ground-based Raman lidar systems have
different overlap functions [Wandinger and Ansmann,
2002], and we obtain different minimum heights above
which trustworthy particle extinction profiles can be derived.
We can largely compensate for this overlap effect by com-
bining the profiles of the three lidar systems. The airborne
HSRL allows us to cross-check the quality of the profiles
acquired by the ground-based lidar measurements at 532
and 1064 nm, as the profiles reach down to the ground.
 Figure 1 illustrates the measurement situation. Time-
altitude plots of the range-corrected signals show the vari-
ability of the dust layer height on these four days. We
selected the wavelength at 710 nm as it is comparably sen-
sitive to the presence of dust particles. We also determined
extinction profiles from measurements of Raman signals
taken during the measurement times shown in Figure 1. The
extinction coefficient profiles are shown on the left part of
each time-altitude plot. We also show mean values of the
particle extinction coefficient measured with MULIS and the
Case Studies: 19 and 28 May 2006,
airborne HSRL, and the column-mean extinction coefficient
that we derived from the AERONET Sun photometer mea-
surements. The measurements were carried out either before
or after the aircraft overflights. The scaling height of Sun
photometer optical depth was determined with lidar signals
taken closest in time to the Sun photometer observations.
 The profiles serve as an indicator whether it is rea-
sonable to compare data products from Sun photometer
(column-averaged and column-integrated data products),
lidar (vertical profiles), and airborne in situ measurements
(carried out in various height levels). It is quite clear that a
match of intensive and extensive parameters is less likely to
occur for the different measurement platforms, if observation
geometry and/or observation time differ, see also the dis-
cussion by Müller et al. [2010a, 2010b]. Regardless of the
proximity of BERTHA and MULIS, and the fact that the
airborne HSRL passed over the Ouarzazate field site it is
obvious that all three instruments sounded different portions
of the dust plume on the four measurement days. An
extinction coefficient that changes with height is a strong
indicator of vertically varying extensive properties.
 Extinction coefficients were measured with BERTHA,
MULIS, and HSRL at 532 nm. We find a good agreement
for the extinction coefficients to heights as low as 1.5 km asl
on 19 May 2006. Differences are at maximum 20% in the
height interval around 4 km height and otherwise consider-
ably less. Regarding the measurement on 28 May 2006 we
find good agreement for the profiles above 3 km height asl.
We find deviations of less than 15% in the plume between
2.8 and 3.6 km height. Below 2.4 km height we could not
measure extinction coefficients with BERTHA. On 3 June,
the extinction values measured with the three instruments
scatter by as much as 50% in each height level. Nevertheless,
we still see similar structures in the extinction profiles, i.e., a
dust layer with increased extinction values reaches up to
approximately 3 km height asl. On top of this layer we find
another layer with lower extinction coefficients up to 4.5 km
asl. Regarding the measurement on 4 June 2006 we find
comparably good agreement between the profiles measured
with MULIS and HSRL. The extinction coefficients mea-
 Note that only BERTHA and HSRL measured
extinction profiles on 3 June. The profiles from MULIS are
obtained by the Klett solution of the backscatter coefficient
[Tesche et al., 2009]. The height-dependent lidar ratio used
in the transformation from profiles of the backscatter to
profiles of the extinction coefficients seems to be chosen
improperly in the upper part of the dust layer. The HSRL
profile furthermore represents an average over 5 s only,
compared to 40, 100, and 13 s averaging intervals on 19
May, 28 May, and 4 June, respectively.
 We also show the time series of optical depth mea-
sured with the AERONET Sun photometer and the extinc-
tion-related Ångström exponents that can be inferred from
the optical depth data. Regarding the extinction-related
Ångström exponent we need to consider that lidar mea-
sures particle extinction at 355 and 532 nm, whereas the
AERONET measurement wavelengths nearest to the lidar
wavelengths are 441 and 501 nm. Thus, any comparison will
suffer from this discrepancy in measurement wavelengths.
A second order fit of the logarithm of aerosol optical depth
versus the logarithm of the measurement wavelength is the
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
4 of 25
Table 2. Measurement Days and Measurement Times (in UTC) of the Different Instrument Platforms That are Used for Comparison Study in Section 3a
Parameter: From Lidar and Sun Photometer According to Table 1
Parameters: ext, scat, abs,
å, psd, criNP, Instrument:
In Situ (Airborne),
Parameter: criCA, Instrument:
In Situ (Airborne),
Lidar: Measurement Time
 BERTHA: 09:59–11:16
 11:23:44–11:37:44: 4853 m (L02)
 11:24:36–1:28:30: 4555 m
 MULIS: 11:04–11:14
 11:44:02–12:06:28: 3247 m (L03)
 11:24:36–11:45:00: 4555 m
 POLIS: 10:59–11:19
 11:08 (40 s)c
(with descent to 3040 m)
comparison time: around 11:00
 BERTHA: 09:53–13:15
 12:42:42–12:49:47: 3870 m (L10)
 11:27:50–11:30:50: 3035 m
 MULIS: 10:37–10:49
 12:52:02–12:54:02: 2907 m (L11)
 11:41:53–11:46:00: 1815 m
 POLIS: in Tinfou
 10:46 (100 s)
 12:55:57–12:59:07: 2194 m (L12)
 11:52:00–11:56:10: 1212 m
comparison time: 10:41–10:43
 BERTHA: 03:56–04:26
 03:39:53–03:44:53: 3798 (L04)
 08:20:07–08:24:30: 3648 m
 MULIS: 04:12–04:16
 03:46:41–03:56:45: 2855 (L05)
 08:31:30–08:34:30: 2428 m
 POLIS: 04:12–04:16
 09:01:43–09:03:04: 2487 (L07)
 08:45:30–08:49:00: 1127 m
comparison time: 04:00–05:00
 03:15 (5 s)  04:14 (5 s)
 09:04:16–09:05:04: 2356 (L08)
(ascent to 1172 m)
 BERTHA: 09:24–12:25
 10:08:18–10:10:02: 3854 (L03)
 11:45:50–11:48:20: 1821 m
 MULIS: 08:53–13:15
 10:12:26–10:20:02: 2895 (L04)
 11:41:45–11:45:50: 3439 m
 POLIS: 09:24–11:25
overpass time: 09:48 and
 10:24:14–10:31:22: 1928 (L05)
(descent to 1832)
comparison time: 09:24–10:29 and 10:40–11:34d
 10:47 (10 s)
aEach data set is denoted by a number in brackets. These numbers are used to identify symbols shown in the figures of this section. Overpass time denotes the time at which the Falcon aircraft passed over the field
site in Ouarzazate. Parameter abbreviations: ext, extinction coefficient; scat, scattering coefficient; abs, absorption coefficient; å, Ångström exponent derived from extinction, scattering, and absorption coefficients that
were computed from particle size distributions and complex refractive indices; psd, particle size distribution; criNP, complex refractive index derived from measurements with nephelometer and PSAP according to
Petzold et al. ; and criCA, complex refractive index derived from chemical analysis of individual particles according to Kandler et al. . We also determined criCAfrom ground-based measurements at
Tinfou; see Figure 3 and explanation in the text.
bThe flight level is given in meter above ground. The numbers in brackets refer to Table 4 of Weinzierl et al. .
cNumbers in brackets of overpass times denote duration of overpass, i.e., data collection time in seconds (s).
dMeasurement times are given for BERTHA. Measurement times were 09:51–10:10 UTC and 10:55–12:15 UTC for MULIS and 09:24–10:16 UTC and 10:30–11:25 UTC for POLIS.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
5 of 25
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
6 of 25
most accurate way to interpolate aerosol optical depth data
to different wavelengths. With respect to the lidar data we
think that such a sophisticated interpolation to the nearest
measurement wavelengths of Sun photometer at 441 and
501 nm will not significantly improve our results. We have
only data at two measurement wavelengths, and the mea-
surement uncertainty of the lidar extinction coefficients is
considerably larger than the accuracy of the optical depth
measurements with Sun photometer.
 For the comparison we furthermore prefer to extrap-
olate the optical depth data taken with Sun photometer at
441 and 501 nm to the lidar measurement wavelengths. We
selected this procedure for consistency reasons. For some of
the lidar data products, i.e., particle lidar ratio and linear
depolarization ratio the lidar instruments did not always
provide us with results at 355 and 532 simultaneously.
Therefore, using extrapolation of the Sun photometer data at
least allows us to do a comparison even if the lidar data
products are available at one measurement wavelength only.
 Figure 1 shows that optical depth and Ångström
exponent did not change significantly during the measure-
ment periods covered by Sun photometer and the lidar
instruments (19 May, 28 May, and 4 June 2006). On 3 June
2006, optical depth slightly decreased between the nighttime
lidar observations and the Sun photometer observations.
 The mean value of the 440/870-nm Ångström expo-
nent was 0.12 on 4 June 2006. The standard deviation
of ?0.04 follows from 46 measurements taken between
0610 and 1542 UTC. The variation of the Ångström expo-
nent was less on the other three days we use for our
comparison. We find values of 0.22 ? 0.02 on 19 May,
0.1 ? 0.01 on 28 May, and 0.11 ? 0.04 on 3 June 2006.
We assume that the Ångström exponent did not change
significantly between nighttime and daytime observations
on all four measurement days. We also see from Figure 1
that the Ångström exponent remained comparably stable
 We find that optical depth from HSRL is lower by
approximately 20% (19 May), 19% (28 May), and 18% and
24% (9:48 UTC and 10:47 UTC on 4 June) compared to
optical depth from Sun photometer. Optical depth measured
with BERTHA is lower than optical depth measured with
the Sun photometer, too. Aerosol optical depth from the lidar
measurements is determined by integration of the profile
of the extinction coefficient. The lower values certainly
cannot be explained by the difference in measurement
wavelengths (501 nm for Sun photometer versus 532 nm for
lidar), given the small wavelength dependence of particle
extinction (see Figure 5). The parameter can show a low bias
toward columnar AERONET measurements, when aerosol
layers are missed in the integration process. This is the
case, e.g., when extinction profiles do not extend all the way
down to the surface as is the case for the profiles from the
ground-based lidars shown in Figure 1. Extinction profiles
from BERTHA are affected by the overlap effect while the
ones from airborne HSRL measurements are very sensitive
to the ground signal. The lowest point of the extinction
profile is determined by the highest topographical feature
within the averaging interval. Another source of bias is ele-
vated layers which are not captured in the lidar measure-
ments but included in columnar AERONET observation.
The latter effect is negligible for SAMUM observations
because the aerosol was restricted to the dust layer. Subvis-
ible cirrus may pose a problem to an accurate retrieval of
dust optical depth [Huang et al., 2011]. Our analysis of the
ground-based and airborne lidar data does not indicate that
subvisible cirrus is responsible for the observed differences.
Furthermore, our ground-based and airborne lidar observa-
tions do not indicate that aerosol layers above the dust layers
contributed to total optical depth in a way that could explain
the systematically larger aerosol optical depths from Sun
photometer compared to optical depth from lidar.
 In summary, the information from our lidars and
the Sun photometer leads us to the conclusion that the dust
plumes were comparably stable (temporally) regarding par-
ticle intensive properties like effective radius, complex
refractive index, and single-scattering albedo. This stability
of the dust plumes is important for our purpose of testing
the quality of some dust data products.
 The particle extinction coefficients varied with height
(except on 19 May 2006). We need to consider this vertical
variation in our comparison of extensive optical and micro-
physical particle properties measured by aircraft (in specific
flight levels), by lidar (vertically resolved, respectively
column-integrated), and by Sun photometer (column-mean
and column-integrated data products).
 Figure 2 shows particle size distributions for the four
measurement days. The results for 28 May and 3 and 4 June
2006 corroborate the findings for 19 May 2006 [Müller et al.,
2010a], i.e., particle size distributions from AERONET Sun
photometer retrievals are different to the results from the
Falcon in situ measurements.
 Inhomogeneities caused by variations of particle
number concentration may be responsible for some of the
observed differences. For this reason we scale our data for
some of the following comparisons. In this way we avoid
Figure 1. Measurements with ground-based lidar (BERTHA) and Sun photometer, and airborne HSRL on (a, b) 19 and
28 May 2006, and (c, d) 3 and 4 June 2006. We show for each of the four measurement days: (1) Time series of particle
optical depth at 501 nm (green dots) and particle extinction-related Ångström exponent (black circles) for the wavelength
pair 441/869 nm from Sun photometer, optical depth from BERTHA (green line), and optical depth from HSRL (orange
triangle); (2) profiles of the particle extinction coefficient at 532 nm from BERTHA, MULIS, HSRL, and mean extinction
coefficients from the AERONET Sun photometer; and (3) the range-corrected backscatter signal at 710 nm measured
with BERTHA. The measurement times of the AERONET Sun photometer are indicated by the vertical dashed lines.
The overflight times of the HSRL are shown by the vertical solid line. The heights in which the in situ data were taken
are shown by the horizontal magenta lines. The start and end of the horizontal lines shows the start and stop of the airborne
measurements. AOT denotes aerosol optical depth and AE denotes the Ångström exponent (computed from optical depths
respective extinction coefficients).
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
7 of 25
the influence of variations of number concentration. Regard-
less of the scaling procedure that we carry out in the follow-
ing, see section 3.2 and Figure 4, we see that the volume
concentration of particles above 5 mm radius measured aboard
the Falcon aircraft is on average larger than the particle vol-
ume concentration inferred by AERONET Sun photometer.
 The AERONET results do not show any particles with
radius above 15 mm radius. This radius of 15 mm is the
maximum particle size considered in the inversion of the
AERONET optical data into particle volume size distributions.
 Regarding the measurement on 19 May 2006 we find
that particle number concentration is less than 0.01 cm?3in
the radius range above 15 mm. We find similar concentra-
tions for 28 May, 3 June and 4 June 2008 [see Weinzierl
et al., 2009, Figures 8 and 9]. The difference of the parti-
cle size distributions to the in situ measurements therefore
cannot be explained by a significant concentration of parti-
cles above 15 mm radius.
 Reid et al. [2003, 2008] discuss in detail error sources
that may lead to errors in particle size distribution measure-
ments with aerodynamic and optical particle sizing instru-
ments. Results impressively show significant variations of
dust size distributions from North Africa and southwest Asia
(Arabian Peninsula). In particular, optical measurement
methods seem to be affected by uncertainties in data analysis.
 The cut point of the inlet and distribution system is
particularly critical for measurements of mineral dust, as the
concentration of coarse mode particles is naturally high.
Reid et al.  report on a value of 10 mm for the system
they used for their observations. Thus there remains some
uncertainty regarding the ability to fully resolve the coarse
mode fraction of dust, as there are also uncertainties
regarding the inlet efficiency. Regarding the Falcon obser-
vations we have to consider such uncertainties, too.
 We carried out an error analysis regarding the particle
size distributions from Falcon. We wanted to test in how far
a wrong assumption of the complex refractive index can
Figure 2. Particle volume size distributions (particle radius holds for volume-equivalent spherical
particles) derived from AERONET Sun photometer measurements (black squares) and aircraft in situ
measurements (colored lines). The results for Sun photometer are the average of several individual
measurements. The numbers in brackets (black color) in each plot window refer to the individual
measurements listed in Table 2. Regarding the airborne measurements we use the parameterizations
of the size distributions in Table 4 of Weinzierl et al. . The numbers in color refer to the colored
lines (blue, red, green, magenta), respectively, which then can be connected to Table 2 where we list
the information on flight times and flight levels. The thin colored curves (with symbols) follow from
the most reasonable assumption on the complex refractive index which is needed in the retrieval of
the particle size distribution. The thick curves (without symbols) result from an extreme error analysis
of the systematic uncertainty; details are given in the text. The green curves in Figure 2a also show
the particle size distributions measured aboard Falcon on 19 May 2006 (same flight levels as for
blue and red curves) [see Müller et al., 2010a, Figure 1]. These size distributions were obtained with
slightly different values of the parameterized lognormal particle size distributions. Vertical error bars
denote one standard deviation.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
8 of 25
cause the significant systematic off-set between the size
distributions from AERONET and from Falcon at particle
radii above 3 mm. The complex refractive index is a neces-
sary a-priori input parameter in the analysis of the data from
the in situ particle sizing instruments. This value decides on
the transformation from scattering signals, which is the pri-
mary data set of the particle sizing instruments, into particle
size. A wrong value of the complex refractive index will
result in wrong particle sizes and thus wrong particle size
distributions. We varied the imaginary part of the complex
refractive index between 0i and 0.004i. We think that these
values determine the reasonable range of numbers that
should be considered in signal analysis of the in situ particle
 The result of our analysis is shown in Figure 2. The
colored curves with symbols represent the particle size dis-
tributions with our best estimate of the complex refractive
index (particularly the imaginary part). The thick colored
curves (without symbols) represent the most extreme sys-
tematic offset. We assumed an imaginary part of 0i, i.e. non-
absorbing dust, which is highly unlikely. We point out that
the Falcon instrumentation included a PCASP (Passive
Cavity Aerosol Spectrometer Probe)and several FSSPs
(Forward Scattering Spectrometer Probe) for measurements
of the coarse mode fraction. There is a radius range between
approximately 0.5–3 mm diameter in which data from
PCASP and FSSP are available [Weinzierl et al., 2009]. Data
analysis is considered successful if there is a match of the
particle concentrations measured with PCASP and FSSP
in this overlap region. This match can be achieved with a
reasonable choice of the complex refractive index. We found
that for certain choices of complex refractive indices we lose
this agreement in the overlap region. In fact this mismatch
can already be seen in the thick curves of Figure 2.
 In conclusion: Our main result of this analysis is that
in fact the Falcon particle size distributions shift toward
smaller particles, which brings them closer to the AERO-
NET particle size distributions. Still, the overall discrepancy
at particle radii above approximately 3 mm remains.
 Figure 3 shows the results for the real and imaginary
part of the complex refractive index. This parameter was
derived from the analysis of the mineralogy of single parti-
cles collected aboard the Falcon aircraft during flights over
Ouarzazate. The complex refractive index was also derived
from particles collected at Tinfou on these days. The com-
plex refractive indices from the airborne data represent
Figure 3. Mean complex refractive index determined from AERONET Sun photometer data (black sym-
bols). The data represent the individual measurements, see Table 2: [5–8] for 19 May, [17–20] for 28 May,
[32–35] for 3 June and [47–49] for 4 June 2006. Complex refractive indices were derived from analysis
of the mineralogy of single particles collected aboard the Falcon aircraft (dark blue symbols) during
flights over Ouarzazate on the four days [11, 12, 24–26, 40–42, 53, 54]. Complex refractive index was
also derived from particles collected at Tinfou (light blue symbols) on these days. Imaginary parts
(red symbols) were derived from observations of dust with SOAP at Tinfou. One set of numbers follows
from the total absorption coefficients (downward pointing red triangles). The second set of numbers
describes the contribution of pure dust (upward pointing red triangles). Details of the retrieval are given
by Müller et al. .
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
9 of 25
particles with radius <1.25 mm. The ground-based results
at the Tinfou site describe particles of radius as large as
 Imaginary parts were derived from observations of
dust with an optical absorption spectrometer at Tinfou
[Müller et al., 2009]. We determined the imaginary part for
the total absorption coefficient, and for the case in which we
exclude the contribution of soot, i.e., we consider pure dust
only. The mean values of the SOAP data in Figure 3 denote
the most likely values. Müller et al. [2010a, Figure 4] show
for the case of 19 May 2006 uncertainty bars. The uncer-
tainty bars describe extreme values, i.e., maximum and
minimum reasonable value. Details of the retrieval are given
by Müller et al. . The uncertainty bars for the results
of 28 May and 3/4 June 2006 do not differ in a significant
way from the uncertainty bars we obtained for the results on
19 May 2006.
 The results for the measurement on 28 May and
3/4 June 2006 are similar to the findings for 19 May 2006
[Müller et al., 2010a]. The real parts from AERONET
retrievals are lower than the real parts from the airborne in
situ observations. We find at minimum 0.1 lower values for
the real part from the Sun photometer to real part from the
in situ data. This minimum deviation mainly occurs in the
wavelength range between 869 to 1021 nm. Deviations
increase to 0.13–0.15 at 441 nm which is the shortest
measurement wavelength of the AERONET instrument.
We also find an increasing deviation of 0.12–0.16 between
AERONET Sun photometer and in situ based real parts at
1638 nm. However, as we pointed out in the methodology
section (section 2), the calibration of the Sun photometer
channel at 1638 generally is worse than the calibration at
the other wavelengths. Thus, this increasing deviation in
the infrared region cannot be taken as final conclusion on the
Figure 4. (a–d) Particle extinction, (e–h) particle scattering, and (i–l) particle absorption coefficients.
Black symbols show the results from AERONET Sun photometer. The error bars (one standard deviation)
describe the variability of the Sun photometer data. Green squares: Particle extinction coefficients were
measured with POLIS at 355 nm [3, 15, 29] (data on 4 June  are of insufficient quality), with BER-
THA at 532 nm [1, 13, 27, 43], and with MULIS at 532 nm [2, 14, 28, 44]; the numbers in brackets refer
to the measurement times and days of Table 2. We use the combination of extinction data (at 532 nm) of
both instruments (BERTHA and MULIS) in order to reduce the overlap height. The particle extinction
coefficient at 1064 nm is estimated from measurements with HSRL [4, 16, 30, 31, 46]. Particle size dis-
tributions (see Figure 2) and complex refractive indices (see Figure 3) (from in situ measurements aboard
the Falcon) were used for Mie-scattering computations. [9, 21, 36, 50] (blue squares), [10, 22, 37, 51]
(blue circles), [23, 38, 52] (blue crosses), and  (blue stars; results for 3 June). The numbers denote
the measurement times of the particle size distributions. Regarding the complex refractive index we used
the mean value of [11, 12] for 19 May 2006, [24–26] for 28 May 2006, [40–42] for 3 June 2006, and
[53, 54] for 4 June 2006. Results of Mie-scattering computations for the measurement on 19 May 2006
(red dots) are shown for comparison, see Figure 4 in Müller et al. [2010a].
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
10 of 25
accuracy of the retrieval of the real parts with Sun photom-
 Regarding the imaginary part we also find differ-
ences. The differences are particularly large at the short
measurement wavelengths. At 441 nm AERONET Sun
photometer results are 0.0002–0.009 lower than the airborne
results. This deviation is equivalent to 5% lower imaginary
parts on 4 June and 63% lower values on 28 May 2006.
 Figure 4 shows the column-mean extinction coeffi-
cients inferred from optical depth measured with AERONET
Sun photometer. We determined the dust layer height from
the range-corrected lidar backscatter signals taken during or
shortly after the Sun photometer observations. Furthermore
we scaled our data to particle extinction at 1064 nm. This
normalization is of advantage as we avoid in this way the
effect of inhomogeneities of the number concentration in the
dust plumes. This is particulary important for the compari-
son with the airborne results.
 We find good agreement to mean extinction coeffi-
cients from the ground-based Raman lidar observations
(POLIS, BERTHA, and MULIS) at 532 nm. The deviations
are larger at 355 nm.
 Figure 4 also shows extinction coefficients derived
from airborne HSRL observations at 1064 nm. We assumed
lidar ratios of around 55 sr in the retrieval [see also Tesche
et al., 2009; Müller et al., 2010a]. Naturally, due to our
scaling, particle extinction agrees to particle extinction from Sun
photometer. If we compare the absolute values (not shown)
Shape-Independent Optical Parameters
we also find good agreement of the mean extinction from
HSRL to mean extinction from the AERONET instrument.
 We derived the particle extinction coefficients from
light-absorption coefficients measured with the 3-wave-
length PSAP (467, 530, and 660 nm) aboard the Falcon
aircraft and particle scattering coefficients which were
derived from nephelometer observations. Details on how
this was done are given by Petzold et al.  and
Weinzierl et al. . A summary of the main steps is
given by Müller et al. [2010a].
 We also derived extinction coefficients from Mie-
scattering computations [Weinzierl et al., 2009]. We used
airborne measurements of particle size distributions [see
Weinzierl et al., 2009, Table 4] and complex refractive
indices as input information for the computations.
 Müller et al. [2010a, Figure 6] show the comparison
of extinction coefficients at the three wavelengths of the
nephelometer/PSAP method and the Mie-scattering compu-
tations for the measurement of 19 May 2006. We find rea-
sonable agreement of the two data sets. Particle extinction
coefficients were inferred with the nephelometer/PSAP
method at 530 nm for the measurements on 28 May and
4 June 2006. Weinzierl et al. [2009, Table 4] provides us
with the necessary information. There is reasonable agree-
ment in magnitude of the measured extinction coefficients to
the Mie-scattering computations at 530 nm, but disagree-
ment in the wavelength dependence, particularly between
the UV and near-IR.
 For this reason we only show results of Mie-scatter-
ing computations using particle size distributions for the
following discussion. We computed the extinction coeffi-
cients for the wavelength range from 350 nm to 1638 nm.
We scaled (normalized) our results to the extinction coeffi-
cient at 1064 nm.
 We find in general a rather flat extinction spectrum,
i.e., particle extinction does not change significantly with
wavelength. This result has to be expected in view of the
large particles that were measured aboard the Falcon aircraft.
The spectral slope of particle extinction is different from
the spectral slope measured by lidar and Sun photometer.
The absolute values of particle extinction (not shown) from
the Mie-scattering computations are larger than what was
measured by lidar and Sun photometer [see also Müller et al.,
2010a, Figure 6; Weinzierl et al., 2009, Figure 12].
 There may be several reasons for the discrepancies
between lidar/Sun photometer data and the aircraft data.
We tested in various ways [Müller et al., 2010a] if deviations
of the extinction coefficients are caused by the time gap in
observation times, or the fact that 1) lidar delivers vertically
resolved extinction coefficients (which can be converted to
column-mean values), 2) Sun photometer observations pro-
vide optical depth (which subsequently can be converted into
column-mean extinction coefficients if dust layer height
from lidar observations is taken), and 3) airborne measure-
ment describe particle extinction coefficients in various flight
levels. We conclude that like in the case of the measurement
of 19 May 2006, dust plume inhomogeneities cannot be the
main cause of the discrepancy.
 Regarding the Mie-scattering computations it is clear
that there are approximation errors in our computations,
because dust particles are not spherical in shape. It is
impossible to assess the exact approximation errors but we
Figure 5. (EXT.) Particle extinction-related Ångström expo-
nents, (SCAT.) particle scattering-related Ångström expo-
nents, and (ABS.) particle absorption-related Ångström
exponents derived from AERONET Sun photometer measure-
ments (black squares) and computed from particle size dis-
tributions and complex refractive indices (circles). We
computed the Ångström exponents for the wavelength pair
441/869 nm from the data shown in Figure 4. Also shown
is the extinction-related Ångström exponent from lidar (bul-
lets). It has been computed for the wavelength pair 355/532
nm. We reanalyzed data on the basis of nephelometer and
PSAP measurements for the absorption-related Ångström
exponents (diamond). Details are explained in the text.
We also show results from PRIDE (cross) [Bergstrom
et al., 2007]. Error bars denote one standard deviation.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
11 of 25
may estimate this error on the basis of our extinction mea-
surements. If we use the results of Figure 4 we see that
approximation errors may be as large as 25% at 532 nm; see
measurement on 19 May 2006. The minimum deviation
is 10% at 532 nm; see measurement on 28 May 2006.
 We carried out computations of particle extinction for
the case of 19 May 2006, assuming in one case spherical
particle shape and in another case spheroids. In both cases
we used the same particle size distributions. We find a 10 %-
deviation for the extinction coefficients. 10% approximation
error, however, also cannot explain for the deviation of the
Falcon data to the remote sensing data. More importantly,
the extinction values are larger (if we look at the absolute
numbers) when we use spheroids in our computations. The
spectral slope did not change significantly, either.
 The discrepancy of extinction coefficients from
airborne data to extinction coefficients from lidar and Sun
photometer increases with increasing measurement wave-
length, which is particularly obvious if we look at the absolute
values (not shown in the plots). The accuracy of optical depth
measurements with Sun photometer is high. Measurement
errors therefore cannot account for the observed differences.
A wrong assumption of dust layer height, which serves as
scaling factor for converting optical depth into extinction
coefficients cannot explain the discrepancies either.
 Uncertainties of the parameterizations of the mea-
sured particle size distributions can be one reason for the
discrepancies. We used the parameterizations of the mea-
sured particle size distributions as given in Table 4 of
Weinzierl et al. . The rather low wavelength depen-
dence of the particle extinction coefficients requires a
considerably high share of large particles. However, we
emphasize once more: the Sun photometer optical depth
measurements are very accurate. We see a non-negligible
wavelength dependence, which indicates that there may
not be such a strong contribution of coarse mode particles,
as suggested by the in situ measurements.
 Figure 4 also shows particle scattering and absorption
coefficients. A detailed discussion is given by Müller et al.
[2010a] for the measurement of 19 May 2006. In that con-
tribution we already noted the large differences between the
scattering coefficients derived from the AERONET and the
airborne in situ data. Figure 4 shows once more the scatter-
ing coefficients from Figure 6 of Müller et al. [2010a]. Our
new results for 28 May, and for 3 and 4 June 2006 corrob-
orate the findings for 19 May 2006. With decreasing wave-
length there is increasing difference between scattering
inferred from the AERONET data to the scattering coeffi-
cients inferred with Mie-scattering computations from the
 We also find increasing differences of the absorption
coefficients from AERONET Sun photometer and Falcon
aircraft data for all four measurement days. The dis-
crepancies increase with decreasing measurement wave-
length. We consider it unlikely that the shape assumption in
the computations, i.e., spherical particles rather than spher-
oids can fully account for the discrepancies. We need to
keep in mind that we use a wavelength-dependent complex
refractive index in the light-scattering computations. Any
error in this wavelength-dependence transfers into errors
of the wavelength-dependent absorption coefficient. As we
can see from Figure 3 there is a significant difference of
this wavelength-dependent imaginary part of the complex
refractive index between the AERONET data and the Falcon
data. Another error source could be that our methods in
general only provide an effective complex refractive index.
This effective value could depend on the measurement
method, which means that the absorption coefficient might
depend on the measurement method.
 Figure 5 summarizes our results for the optical data
in terms of the Ångström exponents. We computed the
Ångström exponents from the Sun photometer, lidar, and in
situ measurements, respectively, for the visible wavelength
range. Extinction-related values from lidar agree to the
AERONET results. We find 0–0.2 for the extinction-related
Ångström exponent. This range describes the mean condi-
tion on all four measurement days for the wavelength pair
441/869 nm (AERONET Sun photometer) respectively the
wavelength pair 355/532 nm (lidar). The airborne results
also show good agreement to the Sun photometer data. The
airborne measurements cover the wavelength range from
441 to 869 nm.
 The choice of the wavelength pairs is important for
the Ångström exponents. We can infer scattering and
absorption coefficients from AERONET data and from
Mie-scattering computations of the airborne in situ data,
but not from lidar. We find a slight discrepancy for the
scattering-related Ångström exponent at all wavelengths
below 700 nm.
 With respect to the absorption-related Ångström
exponent the differences are significant; see also Müller
et al. [2010a]. Figures 4i–4l show that the absorp-
tion coefficients from Sun photometer and the airborne
observations differ significantly at the wavelengths below
approximately 700 nm.
 The Mie-scattering computations of the absorption
coefficients result in absorption-related Ångström exponents
of approximately 4.5–5.5 (wavelength pair 441/869 nm).
These values are considerably larger than the value of
approximately 1.5 that we find from the AERONET data. The
values from our Mie-scattering computations are anomalously
higher than absorption-related Ångström exponents published
in literature. Bergstrom et al.  and Russell et al. 
report on values of 2.34 for Saharan dust observed in the far-
field of the North African source region. The measurements
were carried out during PRIDE in the Caribbean Sea [Reid
et al., 2003]. Coen et al.  report on values as high as
2 derived from absorption measurements of Saharan dust
observed at the high-altitude station Jungfraujoch in Swit-
zerland. Fialho et al.  report on values of 2.9 for
 Figure 4 shows only the results of Mie-scattering
computations in which we use the airborne particle size
distribution measurements and the complex refractive index
found by Kandler et al. ; see Figure 3. This choice of
the complex refractive index to our opinion may deliver a
reasonable maximum value of the absorption coefficients
across the wavelength range considered in our study. Par-
ticularly the strong wavelength-dependence of the imaginary
part, i.e. the strong increase of this parameter toward shorter
wavelengths may be responsible for the unusually strong
increase of the light-absorption coefficient with decreasing
wavelength, i.e., from around 700 nm down to 441 nm. Note
that the results of the optical properties in which we use the
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
12 of 25
nephelometer and PSAP data is shown for 19 May 2006 in
Figures 4a, 4e, and 4i. We do not believe that the choice
of the wavelength range that has been used for the compu-
tations of the absorption-related Ångström exponent can
fully account for the considerably larger values we find in
 In order to find a lower limit of the absorption coef-
ficient and particulary a lower reasonable value of the
absorption-related Ångström exponent we carried out another
sensitivity study in which we tried to extract the effect of an
incorrectly chosen complex refractive index in our data anal-
ysis. For this purpose we once more analyzed our airborne
data of nephelometer and PSAP measurements of dust scat-
tering and absorption properties according to the scheme
described by Petzold et al.  and Weinzierl et al. .
We modified however our assumption on the imaginary part
in this scheme in such a way that we considered lower
 We find a mean value of 3.7, see Figure 5. The vari-
ability, expressed in terms of standard deviation is 1.1.
The lowest value from an individual measurement is 2.6 ?
0.16 (Falcon measurement on 3 June 2006, flight leg L04;
see Table 2), the largest value is 5.3 ? 1.4 (Falcon mea-
surement on 4 June 2006, flight legs L03 and L04). Only
the lowest individual measurement value that we used for
computing the mean value of all 4 comparison days is in the
upper range of numbers reported by Bergstrom et al. 
and Russell et al. .
 We can merely speculate on the possible reasons for
the differences between our results and the numbers reported
in the literature. We measured pure Saharan dust from a
specific source region in North Africa during SAMUM
2006, whereas the location of the source regions in the other
publications was less well defined. Particles observed in
SAMUM were on average larger than what has been repor-
ted by Reid et al.  for PRIDE which focused on
mineral dust in the far field of the North African dust sour-
ces. We do not know if smaller dust particles may lead to
smaller absorption-related Ångström exponents. We do not
know if different measurement techniques used for measur-
ing the same parameter, i.e., the particle light-absorption,
can lead to different absorption-related Ångström exponents.
 In other words: we have insufficient information to
decide if remote sensing methods that infer the absorption
coefficient and its spectral dependence inherently deliver
results different from in situ observations of the light-
absorption coefficient, regardless of measurement errors. We
need to keep in mind that the light-absorbing components
are not evenly distributed in dust grains. Thus, there is the
question, if this spatial distribution in dust may have an effect
on the overall light-absorption capacity at various wave-
lengths, and we do not know in how far the various mea-
surement methods are sensitive regarding this spatial
 In another step we use the AERONET data to com-
pute particle parameters that depend strongly on particle
shape, i.e., extinction-to-backscatter (lidar) ratios and linear
particle depolarization ratios. These parameters are measured
directly with lidar without knowledge of particle shape.
 This part of our study is particularly important for our
goal of using the AERONET light-scattering model for
our lidar data inversion methodology [Müller et al., 1999a,
1999b]. The equations that are used for the computations of
these two parameters are presented by Müller et al. [2010b].
 Figures 6 and 7 show the results for particle lidar ratio
and linear particle depolarization ratio, respectively. Because
Shape-Dependent Optical Parameters
Figure 6. Particle lidar ratios measured with lidar (green) and computed from AERONET data products
(black). The computations are described by Müller et al. [2010b]. The lidar data are from POLIS (355 nm),
MULIS (532 nm), and HSRL (1064 nm). The data describe averages and are computed from the different
lidar instruments and Sun photometer according to the measurement times listed in Table 2.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
13 of 25
of the complexity of the lidar measurements we did not
obtain the complete set of these two parameters for all four
 We find reasonable agreement for the lidar ratio at
532 nm. Discrepancies are particularly obvious at 355 nm.
The numbers from Sun photometer are significantly larger.
We cannot measure with lidar the lidar ratio at 1064 nm.
We estimated this value [Tesche et al., 2009].
 We compare our results to a study carried out with
AERONET Sun photometer data [Cattrall et al., 2005].
The authors present lidar ratios derived from AERONET
Sun photometer observations in various regions on the
globe. One set of results in this study deals with mineral
dust. The authors infer values of approximately 35 sr for the
site of Banizoumbou (Nigeria) and 38 sr for Capo Verde (Sal
Island, Cape Verde Islands). As in this present study, the
particle backscatter coefficient had to be computed from the
AERONET data products of particle size distribution and
complex refractive index. Errors in the computation of the
backscatter coefficients with the AERONET dust light-scat-
tering model may be responsible for the lower lidar ratios
compared to the lidar ratios we obtained from SAMUM; see
also the discussion in section 4 of Müller et al. . Aside
from this error source another source of uncertainty again is
the fact that Sun photometer measures aerosols in the atmo-
spheric column. Dust over Banizoumbou may experience
some impact from continental particles. Dust over Capo
Verde may be affected with marine particles.
 With respect to the linear particle depolarization ratio
we find 20%–30% higher values from lidar compared to
the Sun photometer data in the visible wavelength range
(500– 550 nm). There is good agreement between lidar and
Sun photometer data at 1064 nm.
 We have few data on depolarization ratios. Though
the lidar data underwent careful quality control there still
remains a lot of uncertainty regarding the spectral slope of
the linear particle depolarization ratio. SAMUM 2008 was
carried out in the Republic of the Cape Verde in 2008
[Ansmann et al., 2011]. This field campaign will provide us
with particle depolarization ratios of higher quality. We shall
analyze these data in a future contribution.
4. Statistical Analysis
 In the previous section we discussed in detail the
results of four measurement days for which we have a nearly
complete set of data from several measurement platforms.
The data were taken comparably close to each other in terms
of distance and in terms of temporal off-set between mea-
surement times of each measurement platform, respectively.
In the next step we analyzed the complete measurement
period, which thus offers us a more statistical view on the
inferred data products. We do not compare the data on a day
by day basis as we do not have aircraft observations at
Ouarzazate on more than a few days. Intensive dust optical
and microphysical properties varied comparably little among
Figure 7. Linear particle depolarization ratio measured at 355 nm (green circle; POLIS), 532 nm (green
circle (MULIS) and green triangle (HSRL)), and 1064 nm (green circle; HSRL). Shown are the results for
(a) 19 May [2–4] in Table 2, (b) 28 May 2006 [14, 16], (c) 3 June [27–31], and (d) 4 June 2006 [44, 46].
Error bars denote the systematic uncertainty [see Freudenthaler et al., 2009, Table 2]. Column-mean
linear particle depolarization ratios were calculated from Sun photometer data at 441, 501, 675, 869,
1021, and 1638 nm. The equation used for the computations is given by Müller et al. [2010b]. We used
the Sun photometer measurement times listed in Table 2. We computed mean values for each day. Error
bars denote one standard deviation.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
14 of 25
the different measurement days, see the SAMUM 2006
special issue in TELLUS, 61 B (2009). Table 3 lists the
number of observation days and the number of individual
data sets that we use for our statistical analysis. We point out
that for some of the parameters we have more samples than
measurement days from AERONET, lidar and Falcon mea-
surements. These data sets (number of samples taken) in
large part describe independent measurements, as they were
done at different times (AERONET) or different flight levels
(Falcon). In the case of lidar there are a few cases in which
we split measurements on specific days into two intervals,
as the aerosol conditions notably changed during the mea-
 Figure 8 shows the mean particle size distribution
inferred from AERONET Sun photometer on 24 days. Also
shown is the mean particle size distribution inferred from the
airborne measurements on 6 days on which the aircraft flew
over Ouarzazate. Particle effective radius of the coarse mode
fraction of mineral dust observed with the AERONET Sun
photometer is a factor two lower compared to effective
radius inferred by the airborne observations carried out in
different flight levels on the different flight days [see
Weinzierl et al., 2009; Müller et al., 2010a].
 Reid et al.  report on considerable variation of
dust particle size distributions measured during PRIDE. The
dust was observed with ground-based impactors (MOUDI
and DRUM) and aerodynamic particle sizers (TSI aerody-
namic particle sizer), and airborne PMS (Particle Measuring
Systems) FSSP-100 and PCASP-100X. One conclusion of
the comparison of the particle size distributions measured
with the in situ instruments was that the different measure-
ment techniques may be accountable for the observed var-
iations. The aerodynamic particle sizer used in that study
was considered the most reliable instrument for particle
sizing. In contrast, the airborne FSSP-100 consistently
oversized particles by more than a factor two [Reid et al.,
2003]. Reasons for the oversizing were low responses to
particles in the diameter range from 3 – 10 mm, insufficient
knowledge of the complex refractive index which is needed
Table 3. Measurement Days and Number of Data Sets From the Various Instrument and Measurement Methods That Were Used for the
Statistical Analysis of the Data Products Discussed in Section 4a
AERONETLidar, GroundFalcon Tinfou, Ground
cri, real (441)
cri, real (532)
cri, real (1021)
cri, imag (441)
cri, imag (532)
cri, imag (1021)
aParticle number distribution is denoted by psd. Complex refractive index is denoted by cri. Depol. denotes the linear particle depolarization ratio of dust.
The single-scattering albedo is denoted by ssa. Numbers in brackets denote the measurement (retrieval) wavelength in nm. Regarding the linear particle
depolarization ratio at 532 nm, we only use the results from MULIS [see Freudenthaler et al., 2009, Table 2]. Results from the airborne HSRL are
listed for 532 nm and 1064 nm in Table 2 of Freudenthaler et al. . Here, we only include the results at 1064 nm.
Figure 8. Mean particle size distribution (red) inferred
from AERONET observations. The particle size distribution
represents the mean value of 60 individual size distributions
(black lines); see Table 3. The particle size distribution
includes only data for which the Ångström exponent for the
wavelength pair 501/1021 nm was less than 0.4 [see Tesche
et al., 2009, Figure 4] and the Ångström exponent for the
wavelength pair 379/501 nm was less than 0.7. For that rea-
son we excluded the measurement days of 11/12 May 2006
and 29 May to 1 June 2006. We furthermore excluded 4
observations for which we could not clearly distinguish the
fine mode from the coarse mode fraction of the size distribu-
tion. Also shown is the mean particle size distribution (green)
measured aboard the Falcon aircraft during overflights over
Ouarzazate. We averaged 11 individual size distributions.
The black vertical line denotes the cutoff radius of the Sun
photometer inversion scheme.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
15 of 25
for particle sizing, and the fact that particles are not
 Reid et al. [2003, Table 1] summarize results for
volume median diameters of Saharan dust. The data repre-
sent observations over Niger (West Africa), Tenerife
(Canary Islands in the North Atlantic off the west coast of
Morocco), the Cape Verde Islands (in the tropical North
Atlantic Ocean off the coast of Senegal), and Puerto Rico
(Caribbean Sea). The first place is close to one of the North
African dust sources, i.e., the Bodélé Depression in Chad.
Tenerife and the Cape Verde Islands are still close to the
North African source of dust. Puerto Rico is in the far-field
of Saharan dust. Reid et al.  report on significant
variations of particle size. Regarding the first three sites
volume median diameters (optical equivalent) are >6 mm
(Niger), >8 mm (Tenerife), and 13 ? 2 mm (Cape Verde
Islands). Regarding the far-field site (Puerto Rico) the
authors find a volume median diameter of 9 ? 1 mm.
 A summary on desert dust observations with Sun
photometers shows that optical inversion methods result in
volume median diameters of 3–7 mm [see Reid et al., 2003,
Table 1]. Various Saharan dust regions are considered.
 In view of the critical discussion given by Reid
et al.  on measurement accuracies of various particle
sizing methods there remains the question if the variations
in particle size are due to differences in Saharan dust
regions, transport times and distances (compare Cape Verde
results with those at Puerto Rico), or the applied measure-
 Reid et al.  also report on volume median dia-
meters of desert dust observed during PRIDE. Values are
3.6 mm (geometric diameter). The data were obtained from
measurements with APS (Aerodynamic Particle Sizer). APS
provides in the first step an aerodynamic volume diameter
which subsequently can be transferred into a geometric
median value. For details on how this was done we refer to
Reid et al. . Reid et al. [2003, Figure 9] furthermore
give an impression on the uncertainties of the computations.
It is interesting to note that there appears to be a secondary
peak in the volume size distribution which could shift the
geometric median diameter toward larger values of 6.5 mm,
though this option was ruled out by Reid et al. .
 Saharan dust was observed during the United Arab
Emirates Unified Aerosol Experiment (UAE2) [Reid et al.,
2008]. Measurements were carried out at an Arabian Gulf
coastal site which experienced dust from several sources
within southwest Asia. Among the instruments were aero-
dynamic particle sizer, a DRUM cascade impactor, and an
AERONET Sun photometer. As in the case of the PRIDE
campaign the authors report on volume median diameters of
approximately 3.5 mm and a standard deviation of 30%. The
median diameter ranges from 3.25–4.56 mm in dependence
of dust source region identified in this field experiment [see
Reid et al., 2008, Table 1]. AERONET Sun/sky retrieval
resulted in volume median diameters (optical equivalent)
of 4.1–6 mm.
 Regarding the AERONET Sun/sky inversions we
find volume median diameters between 2.96–8.86 mm for
the SAMUM 2006 data set. The mean value is 4.37 ? 0.95;
we only report the radius for the coarse mode fraction.
Regarding the Falcon in situ measurements, Weinzierl et al.
 find a volume median diameter of 15.5 ? 10.9 mm,
mean values range from 8–18 mm, except for some outliers
[see Weinzierl et al., 2009, Figure 13]. This range of
numbers is considerably larger than the diameters from the
AERONET measurements and the values reported by Reid
et al. [2003, 2008], except for the Cape Verde field site
(13 ? 3 mm). We can only speculate on the reasons for the
 On the one hand measurement errors of the instru-
ments operated aboard the Falcon aircraft may be respon-
sible, aside from the possibility that each of the field sites
(in Niger, Tenerife, Cape Verde Islands, Puerto Rico,
Arabian Gulf region) may have been affected by dust
emitted from different source regions. On the other hand,
Weinzierl et al.  point out in their discussion of the
airborne dust size measurements that the particle number
concentrations measured with the FSSP-300 and the
PCASP-100X showed close agreement in the particle size
region were both instruments measure. This latter argument
may serve as argument that the FSSP observations which
largely decide on our finding of discrepancy between
AERONET and airborne in situ observations may not be the
primary cause of the discrepancies. We must keep in mind
that SAMUM 2006 was a field campaign in which we
obtained comparably strong instrument signals due to the
proximity of our instrument platforms to the source region.
In contrast, PRIDE may be considered as a representative
site for the far-field region. It is still unclear what may
happen to dust size distributions along their journey of sev-
eral thousand kilometers across the Atlantic Ocean. How-
ever, UAE2was also located in a region of strong dust
emissions, still particle size distributions also result in con-
siderable smaller median volume diameters compared to the
results inferred from the Falcon observations; the inlet cutoff
diameter was at 10 mm.
 Figure 8 shows that there is a small fraction of parti-
cles in the size range around 100 nm in radius. We believe
that this fine mode fraction may not necessarily be a retrieval
error, but a real feature of the mean particle size distribution.
We need to keep in mind that the field site is located in the
Draa Valley which is a comparably densely populated
region. There is a significant amount of car traffic, and
emissions from trucks. There are plenty of agricultural
activities, and the burning of agricultural waste is quite
common in this area. Thus there is the possibility that small
particles may be present in the atmospheric column. The
airborne measurements do not cover the lowest few hundred
meters of the atmospheric column, and thus it is not possible
for us to verify this small mode fraction by the airborne
measurements at Ouarzazate.
 Figure 9 presents a comparison of the wavelength
dependence of the real and imaginary parts that are derived
from the AERONET data and from the mineralogical anal-
ysis of the ground-based particle measurements at Tinfou.
The airborne observations of particle complex refractive
index do not differ significantly from the observations at
Tinfou [Kandler et al., 2009].
 The in situ measurements show that the real part is
nearly independent of the measurement wavelength. The
technique of determining the refractive index from mineral-
ogy is based on fixed values for pure minerals, which have
mostly very similar refractive indices and are known with
high accuracy. The variability reflects only changes in
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
16 of 25
composition. The standard deviation of the real part that
follows from the available time series at Tinfou [Kandler
et al., 2009] is between 0.004 and 0.012 (relative standard
deviation of 0.3% to 0.8%), depending on wavelength and
particle size. Including an additional potential variability in
silicate composition, which cannot be identified by our sin-
gle particle analysis, we estimate an uncertainty of 0.03 for
these refractive index values.
 The imaginary part increases with decreasing wave-
length, and we obtain significantly larger values at near
ultraviolet wavelengths. The frequency distribution (describ-
ing the three wavelengths at 441, 532, and 1021 nm) shows
that the overall variability of the imaginary part was rather
low during the measurement period.
 The uncertainty in imaginary part that is derived from
the mineralogical analysis is considerably higher, reflecting
our poor knowledge on basic mineral absorption and the
uncertainties of the mixing rules, as well as the influence of
minor compounds (mainly iron compounds) inside single
particles. At Tinfou, a relative standard deviation between
20% and 50% has been found. Including the mentioned
uncertainties in the base data, the total uncertainty in the
imaginary part is a factor of 2 for dust particles. For a more
detailed discussion we refer to Kandler et al. .
 Petzold et al. [2009, Table 3] report on mean real and
imaginary parts for different dust episodes. The results were
derived from measurements of particle size distributions and
PSAP observations aboard the Falcon aircraft. Mean values
of the real part are between 1.546 and 1.565 for the wave-
lengths 450, 550 and 700 nm, which is close to the values
reported from the mineralogical analysis. Mean imaginary
parts vary between 0.0031–0.0052 at 450 nm and between
0.0003–0.0025 at 700 nm. These numbers are lower than
what we obtain from the mineralogical analysis. The uncer-
tainty is approximately ?25%. We note that these numbers
describe only particles with diameter below 2.5 mm.
 The Sun photometer observations show a compar-
ably weak wavelength dependence of the real part. The
variation of the values at the three wavelengths is broader
compared to the values inferred from the in situ data
(mineralogical analysis). This scatter may be caused by
measurement errors of the optical input data that are used
for data inversion.
 Measurement errors may result from calibration
errors. We cannot exclude that dust optical properties varied
stronger during the complete period of SAMUM 2006 than
what was measured during the comparably few overflight
days of the Falcon aircraft. These changes of optical prop-
erties may either be caused by changes of the particle size
Figure 9. Frequency distributions of the real part and imaginary part of the complex refractive index
determined from AERONET Sun photometer data (grey columns) and from mineralogical analysis of
particles collected at the ground at Tinfou (open columns). Results are shown for the wavelengths at
(a, b) 441 nm, (c, d) 532 nm, and (e, f) 1021 nm. Results of the Tinfou data do not differ in a significant
way from the results of airborne sampling. The results are the mean values. The number of individual
measurements used for this comparison is listed in Table 3.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
17 of 25
distribution or changes in chemical composition. Uncer-
tainties of the inversion procedure may be responsible for a
larger scatter of the results, too.
 Regarding the imaginary part we see that particle-
light absorption increases with decreasing measurement
wavelength. Similarly to the real part, the absolute values
of the imaginary part are different from the values that we
obtain with our in situ measurement technique. On average,
we find lower imaginary parts at the shorter wavelengths
compared to the imaginary parts that follow from the in situ
 One important parameter derived from the SAMUM
2006 campaign is single-scattering albedo. Figure 10 shows
the results. Müller et al. [2010a] discuss in detail the results
of 19 May 2006.
 We used the particle size distributions and complex
refractive indices from the inversion of the Sun photometer
data, and we determined single-scattering albedo at 350,
450, and 550 nm. For this purpose we first computed single-
scattering albedo at 441, 501, 675, 869, and 1021 nm. We
then extrapolated on the basis of a simple linear regression
to 350 nm and 450 nm (we use the results at 441 nm and
501 nm). Regarding the results at 550 nm we use the single-
scattering albedo at 501 and 675 nm for the linear regres-
sion. We assume that the error that is introduced by this
simple extrapolation procedure does not introduce additional
significant uncertainties at 450 and 550 nm. Additional sig-
nificant uncertainties may be introduced at 350 nm. Because
we do not have the complex refractive index from AERO-
NET at this wavelength we have no means of quality
checking. Light-absorption may strongly increase with
decreasing wavelength, and for this reason we may overes-
timate single-scattering albedo from AERONET at 350 nm.
 We obtained the single-scattering albedo from mea-
surements with SOAP [Müller et al., 2009]. SOAP provides
us with the imaginary part of the complex refractive index.
This parameter is given as wavelength dependent quantity.
The details of the retrieval procedure are found in the work
of Müller et al. [2010a].
 We find rather good agreement of the data from
Sun photometer and SOAP at 550 nm. As in the case of all
previous parameters, we again see a difference at the
Figure 10. Frequency distributions of single-scattering
albedo at (a) 350 nm, (b) 450 nm, and (c) 550 nm. Results
are shown for AERONET Sun photometer (grey columns)
and for SOAP measurements (open columns). The results
represent the number of samples listed in Table 3.
Figure 11. Frequency distributions of lidar ratios at
(a) 355 nm, (b) 532 nm, and (c) 1064 nm. Results are shown
for AERONET Sun photometer data (grey columns) and
lidar measurements (open columns). The results represent
the mean of the lidar ratios of the individual samples;
38 samples at 355 nm, 39 samples at 532 nm and 47 samples
at 1064 nm.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
18 of 25
ultraviolet wavelength at 350 nm. SOAP delivers lower
values for the single-scattering albedo than Sun photometer.
 Figure 11 summarizes the lidar ratios measured with
lidar. Lidar ratios at 355 nm are similar to the lidar ratios at
532 nm. Most of the values range between 50 and 60 sr.
Values in the range from 40–50 sr are second most. We do
not have measurements of lidar ratios at 1064 nm.
 The lidar ratios were computed from the Sun pho-
tometer data products at 355, 532, and 1064 nm. We used
the particle size distributions and complex refractive indices
of the Sun photometer retrievals. For the computations we
assumed spheroids and the axis ratio distribution used by
AERONET [Dubovik et al., 2006]. The lidar ratios at
355 nm are significantly larger than the measured lidar
ratios. Most values range from 70–100 sr, compared to 40–
60 sr obtained with lidar. At 532 nm wavelength this dif-
ference becomes less pronounced. The lidar ratio distribu-
tion inferred from the Sun photometer data at 1064 nm
wavelength is rather similar to the one at 532 nm.
 There may be several causes for the deviation
between lidar and Sun photometer results. A high imaginary
part pushes lidar ratios toward higher values. We may
hypothesize that the imaginary part inferred from the
AERONET data is too large at ultraviolet wavelengths. This
assumption however is in contradiction to the fact that the
mineralogical analysis of particles suggests an even higher
imaginary part at ultraviolet wavelengths; see Figure 3.
 We extrapolated the lidar ratios to 355 nm. We
applied a linear fit and we used the lidar ratios at 441 and
501 nm. As in the case of the single-scattering albedo, our
extrapolation procedure that uses the lidar ratios at 441 and
501 nm may introduce error, because the light-absorption
may significantly increase at the shorter wavelengths below
441 nm. However, in this case we see that the AERONET
results of lidar ratio are already significantly larger at
355 nm than what we measured with lidar. So in this sense,
we obtain an opposite behavior of a possibly larger light-
absorption capacity of mineral dust at ultraviolet wave-
lengths. The Sun photometer values may be high in the
case of single-scattering albedo, but they are in fact already
too high for the lidar ratio.
 Small particles can be another reason for larger lidar
ratios. We hypothesize that larger particles, i.e., a larger
share of particles in the coarse mode fraction of the dust size
distribution could lower the lidar ratio. However, this
assumption would be contradictory to the extinction-related
Ångström exponents from Sun photometer and lidar, which
do not show a neutral spectral dependence of particle
extinction. We point out once more that the Sun photometer
observations suggest a factor two smaller particle effective
radii compared to the in situ observations.
 The assumption of size-equivalence in the Sun pho-
tometer inversion algorithm may be another reason that can
lead to the discrepancies of the lidar ratio. It may be wrong
to assume that particles of same size, but different shape
can lead to similar optical properties. This assumption may
be particularly wrong for parameters that strongly depend
on particle shape. The lidar ratio, which describes the
extinction-to-backscatter ratio is such a parameter. Particle
backscattering depends strongly on particle shape. Thus
the use of spheroids, the way it is employed in the Sun
photometer inversion algorithm, may cause discrepancies
between the computed lidar ratios and the measured lidar
ratios. In that regard, also the particle-shape distribution,
respectively the aspect ratio distribution may be a reason
for the observed discrepancies.
 Another reason for the discrepancies of the lidar ratio
is that desert dust aerosol is a mixture of absorbing and non-
absorbing components [Kandler et al., 2009], but only a
single component is considered in the model used by the Sun
photometer inversion algorithm. As Gasteiger et al. 
demonstrate, the non-absorbing components have strong
influence on the lidar ratio at short wavelengths.
 Figure 12 shows that the particle depolarization
ratios from AERONET scatter from low values around 10%
(at 355, 532 and 1064 nm) to values as large as 25% at 355
and 532 nm and 35% at 1064 nm. The mean value from lidar
at 532 nm is larger than the mean value at 532 nm from Sun
photometer. The mean value from lidar at 1064 nm is similar
to the mean value at 1064 nm from Sun photometer.
 The scatter of values may in part result from mea-
surement uncertainties of the input parameters that are used
for the computation of the linear depolarization ratios.
In part the lower depolarization ratios may be caused by
some contribution from particles in the fine mode fraction of
the particle size distribution.
 The modeled depolarization ratio tends to lower
values than what was measured by lidar at 532 nm. On the
basis of all the information on particle size distributions
measured aboard the Falcon and at ground during the one
month period of SAMUM we have no reason to believe that
the fine mode fraction of the particle size distribution is the
main cause for the lower depolarization ratios that we
inferred from the Sun photometer results.
 However, our comparison lacks from a statistical
robustness. The AERONET model has not been designed for
describing light scattering at 180°. We do not have a suffi-
cient data base regarding the lidar measurements at 355 nm.
At 1064 nm we find similar values for the linear particle
depolarization ratio. The data base is rather sparse at this
 As already noted by Müller et al. [2010b], there
seems to be a maximum of the depolarization ratio at 532 nm
and lower values at the other two measurement wavelengths.
The quality of the lidar measurements of the depolariza-
tion ratio was not optimal during SAMUM 2006. We will
expand our study to particle depolarization measurements
carried out during SAMUM 2008 [Ansmann et al., 2011;
Grob et al., 2011].
 Figure 13 shows the aspect ratios of dust particles
collected during several Falcon overflights over Ouarzazate.
The aspect ratio distribution did not change significantly
during 7 overpasses on 4 measurement days. The aspect
ratio distribution peaks around 1.5. We show for comparison
the aspect ratio distribution used by Dubovik et al. .
The frequency of aspect ratios increases with increasing
aspect ratio, and it stops at the value 3. In contrast, the aspect
ratios measured during the overpasses of the aircraft are as
large as 4. However, aspect ratios above 2.5 occur with rather
low frequency. Kandler et al.  determined the aspect
ratios on the basis of particles lying flat on the substrates. The
values therefore may not be fully representative of the real
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
19 of 25
distribution. In the context we also refer to Reid et al. 
who discuss aspect-ratio measurements.
 We carried out light-scattering computations. We
derived lidar ratios and linear particle depolarization ratios
at 355, 532, and 1064 nm. We used the particle size dis-
tributions and complex refractive indices inferred from the
Falcon measurements and derived from the inversion of
the Sun photometer data. We also used the complex refractive
indices derived at Tinfou. We used all possible combinations
of the particle size distributions (from Sun photometer and
from Falcon) and complex refractive indices (from Sun pho-
tometer, Falcon, and Tinfou) for the computations. Regarding
the imaginary part we have two sets of results for Tinfou. One
set of values follows from single-particle analysis. Another
set of results follows from SOAP [Müller et al., 2009]. With
regard to SOAP we again have two choices. One set of
imaginary parts includes the contribution from soot. Another
set describes the influence of dust, only.
 Figure 14 shows a summary of the computations
for one measurement day. A detailed description will be
given in a future contribution. We used for the modeling of
the optical properties of the single particles the method by
Mishchenko and Travis  which has been supplemented
by the geometric optics approach [Macke and Mishchenko,
1996]. The procedure of calculating the optical properties of
mixtures is described by Gasteiger et al. . The aspect
ratio distribution of Dubovik et al.  was applied in the
calculations. The particle radii from the in situ measurements
were applied as cross-section-equivalent radii, whereas the
size distributions from AERONET are given and applied as
volume-equivalent sizes. In the first step we computed the
parameters for the wavelengths at which we have the complex
refractive indices available, which is not necessarily at 355
and/or 532 and/or 1064 nm. For example, from AERONET
data we have values at 441, 501, 675, 869, and 1021 nm. In
that case we extrapolated with a linear regression to the three
lidar wavelengths using the data of the nearest available
wavelengths, for example 441 and 501 nm in the case of
extrapolation to 355 nm.
Figure 12. Histogram distributions of the linear dust depo-
larization ratios at (a) 355 nm, (b) 532 nm, and (c) 1064 nm.
Results are shown for AERONET Sun photometer data
(grey columns) and lidar measurements (open columns).
The results represent the mean of the number of measure-
ment samples listed in Table 3.
Figure 13. Frequency density distribution of aspect ratios determined from single particle analysis.
The meaning of the symbols is given in the legend. FL means flight level. The numbers denote altitude
in m. All measurement times are given in UTC. Also shown is the mean axis ratio distribution of all indi-
vidual measurements carried out during SAMUM 2006 (bullets). For comparison we show the aspect ratio
distribution that is used in the AERONET retrieval scheme (black squares). We also show the aspect-ratio
distribution measured under laboratory conditions (stars); see Dubovik et al.  for details.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
20 of 25
 The spread of the results gives an impression on the
uncertainties of modeled lidar-relevant optical dust proper-
ties. These uncertainties result from the uncertainties of the
microphysical parameters we derived from the different
measurement platforms. This result in fact points out to a
problem of the field measurements, which may be at the root
of the observed discrepancies. Each platform by itself gives
a consistent set of microphysical and optical properties. But
it may be impossible to mix results on parameters obtained
from different measurement platforms. This means, for
example, it may not be feasible that we use particle size
distribution measurements from AERONET in combination
with refractive indices retrieved from mineralogical analysis
for the computation of, e.g., the lidar ratio.
 The linear particle depolarization ratios and lidar
ratios measured with lidar at 355 nm and 532 nm do not
overlap with the results of the computations in the sense that
the lidar derived values are centered inside the range of all
computed lidarratios andlinearparticledepolarization ratios.
We show the results at 1064 nm for completeness. Our lidar
data set is too small to allow for any trustworthy conclusion
on the quality of the computations at each wavelength.
 The depolarization ratios measured with lidar tend to
be at the upper end of values depicted in Figure 14. The lidar
ratios measured with lidar at 355 nm are lower than what is
found from the simulation studies at 355 nm. Regarding the
lidar ratio at 532 nm the situation looks more favorable. The
simulation results are within the variability measured with
 The aspect ratio distribution has some influence on
the linear particle depolarization ratio. Computations with
spheroids of various size parameters show that the linear
depolarization ratio changes [seeGasteiger, 2011,Figure3.2].
We do not find a correlation in the sense that higher aspect
ratios result in higher depolarization ratios, or vice versa.
5.Summary and Conclusions
 In this third part of our series of quality assurance
tests of the SAMUM 2006 results we compare data of the
Figure 14. Comparison of lidar measurements of linear particle depolarization ratios and lidar ratios to
results from simulations with a light-scattering model [Wiegner et al., 2009; Gasteiger et al., 2011].
Shown are the results at (a) 355 nm, (b) 532 nm, and (c) 1064 nm wavelength. We measured these
parameters with lidar on 19 May 2006. The quality of the linear dust depolarization ratios measured with
lidar at 355 and 1064 nm is low. We indicate this by the grey shaded boxes. At 532 nm we have high-
quality data that allow us to show mean value and standard deviation. For the simulations we use
combinations of particle size distributions (psd) and complex refractive indices (cri) measured with the
different SAMUM platforms. The results represent the number of samples given in Table 3. The combi-
nations are shown in the legend. The meaning of the abbreviations is as follows: psd from means that
the measurement of the particle size distribution was done with the respective platform (AERONET
Sun photometer, aboard the Falcon aircraft or at ground at Tinfou), and cri from means that the complex
refractive index was obtained from the respective measurement platforms mentioned for psd from, respec-
tively. SOAP, dust means means that we used the complex refractive index corrected for the influence of
soot. SOAP, total means that we also considered the contribution of soot to the complex refractive index.
T channels means that particle extinction was inferred by the rotational Raman channels of BERTHA
under daylight conditions [Müller et al., 2003; Tesche et al., 2009].
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
21 of 25
complete measurement period. We compare in detail data that
were collected on four days. On these days nearly every
measurement platform provided data. We also present a sta-
tistical overview on some of the parameters that were observed
by AERONET, Raman lidar, high-spectral-resolution lidar,
and in situ instruments during the one-month measure-
 We find differences that likely are not the result of
different measurement times. For example, the intensive
properties of the dust plumes showed comparably little var-
iation throughout the campaign [e.g., Kandler et al., 2009;
Müller et al., 2009; Tesche et al., 2009; Weinzierl et al.,
2009]. We find systematic differences for particle size dis-
tribution, and thus particle effective radius, complex refrac-
tive index (also the wavelength dependence), and important
optical dust properties like absorption and scattering coeffi-
cients, lidar ratios and depolarization ratios.
 We investigated lidar ratios and depolarization ratios
with a data base that contains scattering properties of dust
particles. This data bank contains results for the parameters
computed from the size distributions and complex refractive
index which were derived with the various instrument plat-
forms. We find a large scatter for lidar ratios and linear
particle depolarization ratios, when we combine the different
size distributions and complex refractive indices. Lidar
measures lidar ratio and depolarization ratio directly. The
results that use AERONET parameters do not coincide with
the results measured by lidar.
 SAMUM offers us deep insight into several open
questions of remote sensing of mineral dust and the retrieval
of optical and microphysical dust parameters. This insight
was in great part achieved through coordinated activities of
airborne and ground-based remote sensing instrumentation
and airborne in situ measurements. SAMUM shows dis-
crepancies among dust parameters which likely have to be
attributed to the fact that none of the employed methods can
derive a complete set of dust parameters by itself. Secondary
data products are computed from primary data products, e.g.,
absorption coefficients follow from measurements of parti-
cle size distributions and complex refractive indices. The
assumptions that have to be made in the retrievals of the
secondary data products generate uncertainties which are
unacceptably large in view of the demands of the modeling
community regarding climate forcing studies.
 We summarize our studies into the following lessons:
 1. A set of particle parameters (microphysical and
optical properties) is consistent in itself as long as we derive
this set of parameters from the same platform. This means:
microphysical parameters (particle size distribution and
complex refractive index) from Sun photometer reproduce the
dust optical properties (optical depth, single-scattering albedo,
scattering and absorption coefficients). Falcon microphysical
parameters reproduce the optical properties measured with
nephelometer and PSAP. Note that desert dust is a mixture of
different minerals, each having a different complex refractive
index. Our comparisons suggest that the “effective” complex
refractive index of such mixtures may depend on the mea-
 2. Regardless of this consistency of parameters
within one measurement platform we observe significant
off-set of the same microphysical and optical parameters
among different measurement platforms. For instance,
complex refractive index from Sun photometer is systemat-
ically lower than the one derived from mineralogical analy-
sis of single particles.
 3. It remains an open question if the wavelength-
dependence of the complex refractive index can be derived
with sufficient accuracy so that it can be used for an accurate
computation of dust optical properties and radiative forcing
studies. We find in part significant differences of the wave-
length-dependencies of the imaginary parts from AERONET
Sun photometer, SOAP, and mineralogical analysis of par-
ticles. We believe that the correct description of particle
shape in data analysis may not be the only reason of the
observed differences. Each measurement method may
deliver a different wavelength dependence, for example
because of different measurement respectively retrieval
sensitivities in the different wavelength ranges these meth-
ods use. Furthermore, mineral dust consists of different
chemical components and the final value for the imaginary
part describes this composition in terms of an effective
value. This effective value again may depend on the meth-
odology, as each method may exhibit stronger detection
sensitivity to specific chemical components that are
contained in the particles, respectively.
 4. It remains an open question if any retrieval of the
imaginary part can be verified with another measurement
method that rests upon a different physical measurement
principle. The statement on the complex refractive index can
be summarized as follows: we still lack in useful methods
that allow us to infer the complex refractive index on the
basis of first principles.
 5. Aircraft measurements of particle scattering and
absorption with nephelometer and PSAP, respectively, are not
designed for inferring these parameters across a sufficiently
broad wavelength range. A significant extension of the mea-
surement wavelength range would be necessary for retrieving
particle extinction coefficients at (near) ultraviolet wavelengths
(strong light-absorption by dust) and near infrared wave-
lengths. Only if this extension is done we could perform a
useful comparison with AERONET Sun photometer and
 6. There is a mismatch of absorption and scattering
coefficients between AERONET Sun photometer and Fal-
con aircraft. Regarding the Sun photometer data we believe
that an insufficient description of the wavelength depen-
dence of these parameters, which in fact is related to an
insufficient description of the wavelength-dependence of the
imaginary part, is responsible for absorption coefficients that
are too low at ultraviolet wavelengths. It would be useful, if
AERONET could provide complex refractive indices at
wavelengths shorter than 441 nm on an operational basis.
This extension could benefit instruments that operate at UV
wavelengths like lidar. We are aware that this wavelength
extension may not be straightforward to achieve. Regarding
the airborne measurements it is clear that the measurement
wavelength of around 450 nm (nephelometer, PSAP) is
insufficient to derive absorption coefficients at shorter
wavelengths, which would be urgently needed.
 7. We lack in powerful methods that allow us to
obtain a reasonable link between retrieval results (optical
and/or microphysical parameters) from remote sensing data
and results from so-called in situ observations across suffi-
ciently wide ranges of measurement wavelengths. We need
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
22 of 25
to employ measurement strategies in field missions that
allow us to test if the different methods provide similar
(same) results across a wide range of wavelengths from the
near ultraviolet region to the near infrared region. We need
to learn more about measurement and retrieval uncertainties
and how the uncertainties change with, for example mea-
surement wavelength. In this context we point out to pub-
lications from, e.g. Leahy et al.  and Johnson et al.
 who show reasonable agreement of single-scattering
albedo of smoke and mixtures of smoke and dust observed
with Sun photometer and in situ instrumentation. Toledano
et al.  also find reasonable agreement of single-
scattering albedo for mixtures of smoke with dust observed
over the Cape Verde Islands during the Saharan Mineral
Dust Experiment in 2008 [Ansmann et al., 2011]. Single-
scattering albedo of the SAMUM 2006 data which describes
pure mineral dust does not match as well.
 8. We find systematically higher particle volume
concentrations from in situ measurements aboard Falcon
compared to the volume size distributions retrieved with
AERONET. Our uncertainty computations of the in situ size
distributions show that the choice of the complex refractive
index, which is needed in the retrieval of the Falcon particle
size distributions, has impact at particle radii above 2 mm. But
this effect still cannot account for the observed gross differ-
ences. It remains an open issue to what extend particle shape
has influence on the airborne in situ measurements, and this
question needs to be tackled in future field experiments.
We consider this point particularly important in view of
results by Osborne et al.  who also show that air-
borne in situ size distributions of dust particles are shifted
toward larger particle radii compared to AERONET derived
 9. Particle aspect ratio distributions from micro-
scopic observations are different from the distributions used
by the AERONET algorithm. It remains an open issue if the
aspect ratio distribution is a critical factor in the retrieval of
the AERONET data products.
 10. SAMUM cannot answer the question if spheroids
with smooth surface are the best possible choice in retrieving
dust parameters that depend strongly on particle shape, like
lidar ratios and depolarization ratios. However, it is obvious
that any kind of light-scattering computations assuming
spherical particles should be avoided in data analysis of dust
measurements. This particularly applies to in situ measure-
ments where it is still a common approach to use spherical
particle shape in data analysis.
 11. It seems unlikely that dust lidar ratios from Sun
photometer retrievals can be used as trustworthy input in the
analysis of data collected with so-called backscatter lidars.
Even though we find a reasonable match of lidar ratios at
532 nm to the ones measured with Raman lidar, there simply
remains too much uncertainty on the general applicability of
Sun photometer for retrievals of this parameter. There is a
strong mismatch of the lidar ratio measured at 355 nm with
Raman lidar to the lidar ratio inferred from Sun photometer
 12. Lidar measures particle depolarization ratios and
lidar ratios without assumption on particle shape. For this
reason, Raman lidar and high-spectral resolution lidar must
be the method of choice for any verification of these
parameters by any other remote sensing or in situ measure-
 In view of the studies we carried out until now and
summarizing the results of the SAMUM-2006 field cam-
paign (special issue on SAMUM in Tellus, 61 B, 2009) we
need field campaigns that could help in further resolving
the problems connected with the modeling of optical prop-
erties of mineral dust. We need novel instrumentation and
we need to further improve the measurement strategy we
developed during SAMUM. Otherwise open questions, as
for instance which light-scattering model is appropriate
for describing optical properties of mineral dust? will not
 If we want to make significant progress in dust
research with remote sensing methods we must combine
airborne observations of particles with ground-based and
airborne remote sensing instrumentation. Lidar has to be part
of such studies. This method does not require knowledge of
particle shape for measurements of some optical properties
of mineral dust. The overlap range [Wandinger and
Ansmann, 2002] of such lidars must be as close to the
ground as possible. Thus, we also need Sun photometer at
the lidar site as it can efficiently compensate for the overlap
effect, and it delivers high-quality optical depth data, aside
from additional products like microphysical properties. This
Sun photometer should be operated with the measurement
channel at 1638 nm.
 A novel field experiment on pure mineral dust
(or mixtures of dust with other pollution types) should make
use of a new airborne multiwavelength lidar which could
solve many problems that are connected to ground-based
lidar systems. The ideal candidate for such a lidar would be
an airborne multiwavelength high-spectral-resolution lidar
that delivers backscatter and extinction coefficients at sev-
eral wavelengths plus depolarization information at one or
more wavelengths at day- and nighttime [Burton et al.,
2012; Hair et al., 2008; Rogers et al., 2009].
 We also need airborne in situ instrumentation which
must cover a broad range of particle sizes, at least up to
15 mm particle radius. We must measure the fine mode frac-
tion, too, in order to rule out that particle size distribution
measurements are “contaminated” by anthropogenic pollution.
 Particles need to be collected in various flight levels
for subsequent mineralogical analysis and determination of
aspect ratios, the way it was done during SAMUM 2006.
Novel instruments for measurement ofthe state ofpolarization
of light scattered by the airborne dust particles would fill gaps
in our knowledge of optical properties. An airborne version of
SOAP [Müller et al., 2009] would be highly beneficial for
spectrally resolved measurements of particle light-absorption,
coordinated actions among the different platforms.
 We may go much further with our measurement
strategy and request for novel instrument technology. In
another step this airborne multiwavelength high-spectral-
ground-based lidar systems which should either be Raman
lidar (nighttime measurements of extinction only) or high-
spectral-resolution lidar (daytime and nighttime observations
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
23 of 25
of extinction coefficients). In the most ideal case the two
ground-based lidars should also be multiwavelength systems
for backscatter, extinction and depolarization ratio measure-
ments, and at least one system should be able to measure in a
wavelength range larger than the currently used upper limit
of 1064 nm. For example laser wavelengths and detec-
tion channels around 1600–2000 nm could be tackled in
 The first ground-based system should emit the laser
beam(s) vertically. The second ground-based system should
not only emit its own laser beam(s), but this second system
should be operated several kilometers away from this first
ground-based lidar. The most suitable distance depends on
dust layer height over the first ground-based system. The
signal-receiver system of this second ground-based system
must have a scanning unit such that it can scan along the laser
beam that is emitted vertically by the first ground-based
system. The signal-detection unit of this second system must
be able to detect the laser radiation that is scattered to the
sides of the vertically outgoing laser beam of the first system.
Clearly, such an instrument set-up requires sophisticated
alignment and synchronization of the two systems regarding
the laser-pulse emission unit and the signal detection unit.
 In the end we obtain a sophisticated bi-static lidar
set-up that could permit us to measure sideward-scattered
laser radiation. It is clear that we can measure with the
second, scanning system, at each height level of the verti-
cally emitted laser beam laser radiation at one sideward
scattering angle only. However, if the dust plume is high
enough and homogeneous, we will be able to reconstruct
the particle scattering phase function for a specific range
of scattering angles on the basis of radiation measured at
different scattering angles in different height ranges at one
or more wavelengths.
 Finally this second, scanning system, should also
measure the signal of the laser radiation it emits itself.
Particularly along the line of the vertically pointing laser
beam of the first system we thus would also obtain particle
backscatter and extinction coefficients which could serve
as a powerful tool of quality assurance of the overall set of
data collected with these two systems.
 In summary, this idea goes far beyond any field
campaign that has been designed so far. We had plans of
carrying out tests of such instrument configurations during
SAMUM-2006 as some of the lidar instrumentation was
available. The lack of preparation time and the lack of time
during the field campaign did not allow us to do this work.
3112. The SAMUM campaign was funded by the German Research
Foundation (Deutsche Forschungsgemeinschaft) within the Research Group
SAMUM under grant FOR 539. We thank Oleg Dubovik and the
AERONET team at Goddard Space Flight Center for providing us with
high-quality Sun photometer results. We are grateful to the Moroccan
Ministry for Foreign Affairs and the Ministry of the Interior for their
permission to carry out the SAMUM field campaign in Morocco. We thank
the Moroccan Airport Authority and in particular respectable Monsieur
support of the participants of SAMUM.
This study was funded by CATER 2009-
Althausen, D., D. Müller, A. Ansmann, U. Wandinger, H. Hube, E. Clauder,
and S. Zörner (2000), Scanning 6-wavelength 11-channel aerosol lidar,
J. Atmos. Ocean. Tech., 17, 1469–1482.
Ansmann, A., A. Petzold, K. Kandler, I. Tegen, M. Wendisch, D. Müller,
B. Weinzierl, T. Müller, and J. Heintzenberg (2011), Saharan Mineral
Dust Experiments SAMUM-1 and SAMUM-2: What can be learned from
these dust closure field campaigns?, Tellus, Ser. B, 63, 403–429,
Bergstrom, R. W., P. Pilewskie, P. Russell, J. Redemann, T. C. Bond, P. K.
Quinn, and B. Sierau (2007), Spectral absorption properties of atmo-
spheric aerosols, Atmos. Chem. Phys., 7, 5937–5943.
Böckmann, C., I. Miranova, D. Müller, L. Scheidenbach, and R. Nessler
(2005), Microphysical aerosol parameters from multiwavelength lidar,
J. Opt. Soc. Am.-A, 22, 518–528.
Burton, S. P., R. A. Ferrare, C. A. Hostetler, J. W. Hair, R. R. Rogers, M. D.
Obland, C. F. Butler, A. L. Cook, D. B. Harper, and K. D. Froyd (2012),
Aerosol classification system using airborne High Spectral Resolution
Lidar measurements: Methodology and examples, Atmos. Meas. Tech.,
5, 73–98, doi:10.5194/amt-5-73-2012.
Cattrall, C., J. Reagan, K. Thome, and O. Dubovik (2005), Variability of
aerosol and spectral lidar and backscatter and extinction ratios of key
aerosol types derived from selected Aerosol Robotic Network locations,
J. Geophys. Res., 110, D10S11, doi:10.1029/2004JD005124.
Coen, M. C., E. Weingartner, D. Schaub, C. Hueglin, C. Corrigan, S. Henning,
M. Schikowski, and U. Baltensperger (2004), Saharan dust events at the
Jungfraujoch: Detection by wavelength dependence of the single scattering
albedo and first climatology analysis, Atmos. Chem. Phys., 4, 2465–2480.
Dubovik, O., and M. D. King (2000), A flexible inversion algorithm for
retrieval of aerosol optical properties from Sun and sky radiance measure-
ments, J. Geophys. Res., 105, 20,673–20,696.
Dubovik, O., et al. (2006), The application of spheroid models to account
for aerosol particle nonsphericity in remote sensing of desert dust,
J. Geophys. Res., 111, D11208, doi:10.1029/2005JD006619.
Eck, T. F., B. N. Holben, J. S. Reid, O. Dubovik, A. Smirnov, N. T.
O’Neill, I. Slutsker, and S. Kinne (1999), Wavelength dependence
of the optical depth of biomass burning, urban, and desert dust aerosols,
J. Geophys. Res., 104, 31,333–31,350.
Esselborn, M., M. Wirth, A. Fix, B. Weinzierl, K. Rasp, M. Tesche, A. Pet-
zold, and G. Ehret (2009), Spatial distribution and optical properties of
Saharan dust observed by airborne high spectral resolution lidar during
SAMUM 2006, Tellus, Ser. B, 61, 131–143, doi:10.1111/j.1600-
Fialho, P., A. D. A. Hansen, and R. E. Honrath (2005), Absorption coeffi-
cients by aerosols in remote areas: A new approach to decouple dust and
black carbon absorption coefficients using seven-wavelength Aethal-
ometer data, J. Atmos. Sci., 36, 267–282.
Freudenthaler, V., et al. (2009), Depolarization-ratio profiling at several
wavelengths in pure Saharan dust during SAMUM 2006, Tellus, Ser. B,
Gasteiger, J. K. (2011), Retrieval of microphysical properties of desert dust
and volcanic ash aerosols from ground-based remote sensing, diss., 137
pp., Universität München, Munich, Germany.
Gasteiger, J., M. Wiegner, S. Gross, V. Freudenthaler, C. Toledano,
M. Tesche, and K. Kandler (2011), Modelling lidar-relevant optical prop-
erties of complex mineral dust aerosols, Tellus, Ser. B, 63, 725–741,
Grob, S., et al. (2011), Characterization of the planetary boundary layer dur-
ing SAMUM-2 by means of lidar measurements, Tellus, Ser. B, 63, 695–
Hair, J. W., C. A. Hostetler, A. L. Cook, D. B. Harper, R. A. Ferrare,
T. L. Mack, W. Welch, L. R. Izquierdo, and F. E. Hovis (2008), Air-
borne high-spectral-resolution lidar for profiling aerosol optical pro-
files, Appl. Opt., 47, 6734–6752, doi:10.1364/AO.47.006734.
Heese, B., D. Althausen, T. Dinter, M. Esselborn, T. Müller, M. Tesche,
and M. Wiegner (2009), Vertically resolved dust optical properties during
SAMUM: Tinfou compared to Ouarzazate, Tellus, Ser. B, 61, 195–205.
Holben, B. N., et al. (1998), AERONET: A federated instrument network
and data archive for aerosol characterization, Remote Sens. Environ.,
Holben, B. N., et al. (2001), An emerging ground-based aerosol climatol-
ogy: Aerosol optical depth from AERONET, J. Geophys. Res., 106,
Huang, J., N. C. Hsu, S. Tsay, M.-J. Jeong, B. N. Holben, T. A. Berkoff,
and E. J. Welton (2011), Susceptibility of aerosol optical thickness retrie-
vals to thin cirrus contamination during the BASE-ASIA campaign,
J. Geophys. Res., 116, D08214, doi:10.1029/2010JD014910.
Johnson, B. T., S. Christopher, J. M. Haywood, S. R. Osborne, S. McFarlane,
C. Hsu, C. Salustro, and R. Kahn (2009), Measurements of aerosol proper-
ties from aircraft, satellite and ground-based remote sensing: a case-study
from the Dust and Biomass-burning Experiment (DABEX), Q. J. R.
Meteorol. Soc, 135, 922–934.
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
24 of 25
Kandler, K., et al. (2009), Size distribution, mass concentrations, chemical Download full-text
and mineralogical composition, and derived optical parameters of the
boundary layer at Tinfou, Morocco, during SAMUM 2006, Tellus, Ser.
B, 61, 32–50.
Kandler, K., et al. (2011), Ground-based off-line measurements at Praia,
Cape Verde, during the Saharan Mineral Dust Experiment: Microphysical
properties and mineralogy, Tellus, Ser. B, 63, 725–741, doi:10.1111/
Knippertz, P., et al. (2009), Dust mobilization and transport in the northern
Sahara during SAMUM 2006, Tellus, Ser. B, 61, 12–31.
Leahy, L. V., T. L. Anderson, T. F. Eck, and R. W. Bergstrom (2007),
A synthesis of single scattering albedo of biomass burning aerosol over
southern Africa during SAFARI 2000, Geophys. Res. Lett., 34, L12814,
Macke, A., and M. I. Mishchenko (1996), Applicability of regular shapes
in light scattering calculations for atmospheric ice particles, Appl. Opt.,
35, 4291–4296, doi:10.1364/AO.35.004291.
Mishchenko, M. I., and L. D. Travis (1998), Capabilities and limitations of
a current Fortran implementation of the T-matrix method for randomly
oriented rotationally symmetric scatterers, J. Quant. Spectrosc. Radiat.
Transfer, 60, 309–324.
Müller, D., U. Wandinger, and A. Ansmann (1999a), Microphysical particle
parameters from extinction and backscatter lidar data by inversion with
regularization: Theory, Appl. Opt., 38, 2346–2357.
Müller, D., U. Wandinger, and A. Ansmann (1999b), Microphysical parti-
cle parameters from extinction and backscatter lidar data by inversion
with regularization: Simulation, Appl. Opt., 38, 2358–2368.
Müller, D., I. Mattis, U. Wandinger, D. Althausen, A. Ansmann, O. Dubovik,
S. Eckhardt, and A. Stohl (2003), Saharan dust over a central European
EARLINET-AERONET site: Combined observations with Raman lidar
and Sun photometer, J. Geophys. Res., 108(D12), 4345, doi:10.1029/
Müller, D., I. Mattis, A. Ansmann, U. Wandinger, C. Ritter, and D. Kaiser
(2007), Multiwavelength Raman lidar observations of particle growth
during long-range transport of forest-fire smoke in the free troposphere,
Geophys. Res. Lett., 34, L05803, doi:10.1029/2006GL027936.
Müller, D., et al. (2010a), Mineral dust observed with AERONET Sun pho-
tometer, Raman lidar, and in situ instruments during SAMUM 2006:
Shape-independent particle properties, J. Geophys. Res., 115, D07202,
Müller, D., et al. (2010b), Mineral dust observed with AERONET Sun
photometer, Raman lidar, and in-situ measurements during SAMUM
2006: Shape-dependent particle properties, J. Geophys. Res., 115,
Müller, T., A. Schladitz, A. Maßling, N. Kaaden, A. Wiedensohler, and
K. Kandler (2009), Spectral absorption coefficients and imaginary parts
of refractive indices of Saharan dust during SAMUM-1, Tellus, Ser. B,
Osborne, S. R., B. T. Johnson, J. M. Haywood, A. J. Baran, M. A. J.
Harrison, and C. L. McConnel (2008), Physical and optical properties
of mineral dust aerosol during the Dust and Biomass-burning Experi-
ment, J. Geophys. Res., 113, D00C03, doi:10.1029/2007JD009551.
Petzold, A., et al. (2009), Saharan dust refractive index and optical proper-
ties from aircraft-based observations during SAMUM 2006, Tellus, Ser.
B, 61, 118–130.
Reid, J. S., et al. (2003), Comparison of size and morphological mea-
surements of coarse mode dust particles from Africa, J. Geophys. Res.,
108(D19), 8593, doi:10.1029/2002JD002485.
Reid, J. S., E. A. Reid, A. Walker, S. Piketh, S. Cliff, A. A. Mandoos, S.-C.
Tsay, and T. F. Eck (2008), Dynamics of southwest Asian dust particle
size characteristics with implications for global dust research, J. Geophys.
Res., 113, D14212, doi:10.1029/2007JD009752.
Rogers, R. R., et al. (2009), NASA LaRC airborne high spectral resolution
lidar aerosol measurements during MILAGRO: Observations and valida-
tion, Atmos. Chem. Phys., 9, 4811–4826.
Russell, P. B., et al. (2010), Absorption Angstrom Exponent in AERONET
and related data as an indicator of aerosol composition, Atmos. Chem.
Phys., 10, 1155–1169.
Tesche, M., et al. (2009), Vertical profiling of Saharan dust with Raman
lidars and airborne HSRL in southern Morocco during SAMUM, Tellus,
Ser. B, 61, 144–164.
Toledano, C., M. Wiegner, M. Garhammer, M. Seefeldner, V. Freudentha-
ler, J. Gasteiger, D. Müller, and P. Köpke (2009), Spectral aerosol optical
depth characterization of desert dust during SAMUM 2006, Tellus, Ser.
B, 61, 216–228.
Toledano, C., et al. (2011), Optical properties of aerosol mixtures
derived from Sun-sky radiometry during SAMUM-2, Tellus, Ser. B, 63,
Veselovskii, I., A. Kolgotin, V. Griaznov, D. Müller, U. Wandinger, and
D. N. Whiteman (2002), Inversion with regularization for the retrieval
of tropospheric aerosol parameters from multiwavelength lidar sounding,
Appl. Opt., 41, 3685–3699.
Veselovskii, I., O. Dubovik, A. Kolgotin, T. Lapyonok, P. D. Girolamo,
D. Summa, D. N. Whiteman, M. I. Mishchenko, and D. Tanré (2010),
Application of randomly oriented spheroids for retrieval of dust particle
parameters from multiwavelength lidar measurements, J. Geophys. Res.,
115, D21203, doi:10.1029/2010JD014139.
Wandinger, U., and A. Ansmann (2002), Experimental determination of
the lidar overlap profile with Raman lidar, Appl. Opt., 41, 511–514.
Weinzierl, B., A. Petzold, M. Esselborn, M. Wirth, K. Rasp, K. Kandler,
L. Schütz, P. Köpke, and M. Fiebig (2009), Airborne measurements
of dust layer properties, particle size distribution and mixing state of
Saharan dust during SAMUM 2006, Tellus, Ser. B, 61, 96–117.
Wiegner, M., et al. (2009), Numerical simulations of optical properties
of Saharan dust aerosols with emphasis on linear depolarization ratio,
Tellus, Ser. B, 61, 180–194.
D. Althausen, A. Ansmann, D. Müller (corresponding author), and
T. Müller, Leibniz Institute for Tropospheric Research, Permoserstr. 15,
D-04318 Leipzig, Germany. (firstname.lastname@example.org; email@example.com; detlef@
J. Gasteiger, Meteorological Institute, Ludwig Maximilian University,
Theresienstr. 37, D-80333 Munich, Germany. (firstname.lastname@example.org)
K. Kandler, Institut für Angewandte Umweltwissenschaften, Technische
Universität Darmstadt, D-64287 Darmstadt, Germany. (email@example.com)
K.-H. Lee, Atmospheric Remote Sensing Laboratory, School of
Environmental Science and Engineering, GIST, 123 Cheomdan-gwagiro
(Oryong-dong), Buk-gu, Gwangju 500-712, South Korea. (lkh1515@gist.
S. Otto, Leipzig Institute for Meteorology, University of Leipzig,
Stephanstr. 3, D-04103 Leipzig, Germany. (firstname.lastname@example.org)
M. Tesche, ITM, Stockholm University, Svante Arrhenius väg 8, SE-
10691 Stockholm, Sweden. (email@example.com)
C. Toledano, Grupo de Óptica Atmosférica, Universidad de
Valladolid, Prado de la Magdalena s/n, E-47071 Valladolid, Spain.
B. Weinzierl, Institut für Physik der Atmosphäre, DLR, D-82234
Wessling, Germany. (firstname.lastname@example.org)
MÜLLER ET AL.: SAMUM 2006 — COMPARISON STUDY
25 of 25