Site and flight description
During the PEGASOS measurement campaign 2012, vertical profiles were
performed near the San Pietro Capofiume (SPC) ground station located in the
Po Valley in Italy (see Fig. ). The general set-up of the
Zeppelin NT platform for aerosol measurements as well as meteorological data
for the flight on 20 June 2012 were presented in . The SPC
station is a rural background site well suited to investigate aerosols which
have been transported over longer distances. Due to its vicinity to cities
like Bologna (∼ 40 km to the south-west), it also offers the
possibility to study pollution from regional sources. Several campaigns have
already taken place at this site focusing on variations in chemical
composition as well as hygroscopic and optical properties (e.g. ). The station is
equipped with instruments comparable to those employed on the Zeppelin NT
airship with the addition of a nephelometer for a direct measurement of the
aerosol scattering coefficient and a lidar.
In order to get an estimate of the mixing layer height at a certain time an
automated lidar-ceilometer (Jenoptik CHM15K “Nimbus”) was operated at SPC.
In the present analysis we employed an operator-driven approach which avoids
the major drawbacks of automated mixing layer height (MLH) retrievals
(e.g. ). This is performed by
manually evaluating the MLH by a skilled operator's visual analysis of three
plots obtained by the pre-processing of the ceilometer signal (the plots are
presented in the Supplement): (1) the range corrected signal plot; (2) the
signal's gradients plot, and (3) the signal's variance plot (e.g. ).
By observing these three plots, the trained operator
manually marks a number of points on the plots (at least one per hour)
matching the requirements of showing maximum signal gradients, maximum signal
variance, continuity between sunrise until sunset and separation from the
residual layer's gradient maxima. A spline curve is then fitted to these
points to provide a continuous MLH over time. Naturally, the MLH is not
retrieved when it descends below the minimum height observed by the
ceilometer (about 200 m above ground). We found, if the ML aerosol
imprint is present in the signal (as it is in the ceilometer record addressed
in this paper), the indetermination in the MLH retrieval is of the order of
2–3 signal bins (i.e. ±30–45 m). Additionally, we compared the
ceilometer retrieval to the MLH found by analysing T and RH gradients from a
collocated radio sounding performed at 11:00 UTC (the corresponding
figure can be found in the Supplement). Note that the radio sounding was
carried out only once every 12 h, while ceilometer retrievals of MLH have
a time resolution of minutes. The 11:00 UTC MLH retrieved from the radio
sounding yielded a value of 753 m, while an altitude of 772 m
was found from the ceilometer data at this time of day. The two retrievals
agree within the ±45 m we commonly use as uncertainty of our MLH
retrievals. Detailed height profiles of the potential temperature (Θ),
which support the findings of the ceilometer, can be found in the Supplement.
On 20 June 2012 a set of vertical profiles of aerosol particle properties
were obtained between 50 and 800 m near the SPC ground station starting
from the early morning (∼ 08:00 LT) and ending in the early
afternoon (∼ 14:00 LT) with a short refuel break in between
(∼ 10:00 and 11:00 LT). The goal of these flights was to study how the dynamics
of the PBL affects the vertical and temporal variability of the observed and
derived aerosol parameters. The day was characterized by low wind speeds of
approximately 2–3 m s-1 with mainly westerly wind direction.
Therefore, local emissions are expected to have a strong influence.
Instrumentation
In the following we present only those of all PEGASOS instruments used for
this analysis.
Aerosol particle size distributions
To obtain dry aerosol particle size distributions, scanning mobility particle
sizers (SMPS; e.g. ) and a white-light aerosol
spectrometer (WELAS; Palas GmbH, Type 2300; see or
for more details) were used. The WELAS is an optical
instrument that uses a white-light source (OSRAM XBO-75 xenon short arc lamp)
which minimizes Mie oscillations for the light scattering and enables mostly
unambiguous attribution of particle optical diameters to measured scattering
cross sections for most aerosol types. Nominally, the instrument covers the
size range between approximately 200 nm and 10 µm but
its actual range is dependent on the index of refraction of the measured
particles (see below).
At the SPC ground station, an SMPS (custom-built instrument from TROPOS,
Leipzig with a butanol–condensation particle counter) was used to measure
the aerosol particle size distributions for dry particles with diameters
between 10 and 800 nm. The SMPS system was set up in the usual way
that particles were first neutralized in a bipolar particle charger, then
classified according to their electrical mobility in a differential mobility
analyzer (DMA) and finally counted in a condensation particle counter (CPC).
The size distributions were corrected for particles with multiple charges. The
airborne data sets were recorded using an SMPS (TSI Inc., DMA Model 3081 and
water – CPC Model 3786) and a WELAS resulting in a combined dry aerosol
particle size distribution between about 10 nm and
10 µm. The airborne SMPS only measured particle sizes between
10 and 430 nm. The WELAS system recorded particle sizes above
approximately 500 nm, since observations for smaller particles were
discarded, as they were potentially biased by a reduced counting efficiency
. The size range of airborne SMPS and WELAS thus did
not overlap; therefore a spline interpolation was performed in between using
the surface area size distributions measured by the instruments. The
combination of data from both measurements made it possible to cover the full
optically relevant size range. The resulting size distributions were
estimated to have an uncertainty of ±12 and ±5 % for the number
concentrations and the diameters, respectively.
Hygroscopic properties
The airborne platform was equipped with the white-light humidified optical
particle spectrometer (WHOPS) to measure the hygroscopic growth factor (GF),
defined as the ratio of the particle diameter at an elevated RH (Dwet)
to the one at dry conditions (Ddry):
GF(RH)=Dwet(RH)Ddry.
The GF was recorded for dry particle diameters of 500 nm. A detailed
description of the design and specifications of the WHOPS and associated data
analysis procedures was provided in . Briefly, particles
are first dried before quasi-monodisperse aerosol particles with a
well-defined dry diameter (Ddry) are selected in a DMA. The
scattering cross section (“optical size”) of these particles is then
alternately determined at dry conditions and at high RH by either leading the
particles directly into the WELAS or by first exposing them to typically
95 % RH before measurement in the WELAS. The dry responses from the two
different techniques can then be compared to infer the index of refraction of
the selected dry particles (details on the approach are presented in
). Assuming an index of refraction, the scattering cross
section can be calculated from the dry diameter using Mie theory. The index
of refraction that brings this theoretical scattering cross section into
agreement with the measured one, is defined as the effective index of
refraction in the context of this work. The term “effective” reflects the
fact that several simplifying assumptions are made in the Mie calculations.
The particles are assumed to be perfectly spherical and present a homogeneous
internal mixture, and the imaginary part of the index of refraction is
assumed to be zero. The latter approximation is justified by the fact that
scattering coefficients exceed absorption coefficients by a factor of 8 (see
Sect. and Fig. ). In this manner, an average
effective index of refraction of 1.43 ± 0.04 (±1σ uncertainty)
was determined by for the particles probed during the
flight in this study. The humidified mode aims at measuring the hygroscopic
growth factors of the selected particles, following the approach in
. For this purpose, the measured scattering cross section
of the humidified particles is converted to an optical diameter representing
Dwet, such that the hygroscopic growth factor can be inferred with
Eq. (). In order to obtain meaningful Dwet and GF
values, it is crucial to use the true index of refraction of the solution
droplets in the Mie calculations. The index of refraction is obtained as the
volume-weighted mean of the indices of refraction of the dry particles and
pure water (mH2O = 1.333) according to the respective volume
fractions at a certain GF. The relative uncertainty of the inferred
hygroscopic growth factors was found to be approximately ±10 % for
GF < 3 .
The ground station in SPC was equipped with a hygroscopicity tandem
differential mobility analyzer (HTDMA; see e.g. ).
Here, GF for dry diameters of 200 nm were used (compare
). In the HTDMA two DMAs are operated in series and
connected to a CPC. In the first DMA dry monodisperse aerosol particles are
selected and then exposed to a defined elevated RH. The second DMA coupled to
the CPC is used to measure the size distribution of the grown particles. The
uncertainty for these HTDMA-GFs is expected to be approximately
±5 %, if a ±2 % uncertainty is assumed for the RH measurement.
The hygroscopic growth measured with the WHOPS and the HTDMA were used to
convert scattering coefficients obtained from measurements of the dry aerosol
to the corresponding value at ambient RH, as detailed in Sect. .
Aerosol scattering coefficient
At the SPC ground station the total light scattering coefficients were
measured with an integrating nephelometer (TSI Inc., Model 3563) at three
different wavelengths of λ = 635, 525, and 450 nm behind
a PM10 inlet system and after drying to RH < 40 %. The
truncation error correction introduced by was applied.
The uncertainty for these measurements is estimated to be ±5 %. As
no direct measurement of the aerosol scattering coefficient was available
aboard the Zeppelin NT airship, it was inferred using the particle size
distributions, the effective index of refraction and Mie theory assuming
spherical particles . First,
scattering cross sections (σs) as a function of particle
diameter (D) were calculated using the wavelength of
λ =5 20 nm and the range of indices of refraction (m)
measured during the flight. This specific wavelength was chosen to compare
the airborne data to results from the ground-based and remote-sensing
measurements. The WHOPS retrieval yielded on average m = 1.43 ± 0.04, while
a comparison to the directly measured scattering coefficients from SPC showed
that using m = 1.43 ± 0.02 for the Mie calculations is enough to explain the
variability of the nephelometer data. Second, the scattering coefficients (μs) were obtained by integrating the product of
σs and the measured number size distributions
dNdD over the full diameter range:
μs,j(λ,m,D)=∫DminDmaxσsλ,mj,Dj⋅dNdDjdDj.
The index j can be replaced by dry when calculating the dry scattering
coefficients or by wet when the humidified coefficient is regarded. An
uncertainty analysis showed that changes in the index of refraction caused
the biggest errors in μs,dry. Together with the size distribution
uncertainty an overall uncertainty of approximately ±18 % was
obtained for the dry scattering coefficient. It is possible to directly
compare ground based and airborne measurements with each other, as both were
performed at dry conditions.
For comparison with the lidar remote-sensing data, the Zeppelin measurements
were corrected to ambient RH. The importance of this correction was
previously studied by using a humidified nephelometer (Wet-Neph;
).
This instrument directly measures the scattering enhancement due to elevated
RH, which can be described by a wavelength (λ) dependent scattering
enhancement factor f(RH, λ):
f(RH,λ)=μs,wet(RH,λ)μs,dry(λ).
As no such instrument was available during the PEGASOS campaign, the humidity
correction was achieved by combining the GF results at 95 % RH from
with the ambient RH measurements to determine an ambient
light scattering coefficient. This makes it possible to infer the effect of
hygroscopic growth on the light scattering coefficient by considering its
effect on the size distribution. The GF was measured for monodisperse aerosol
particles of 500 nm but we assume it to be constant over the full
size range. Small particles (D < 200 nm) could potentially have
a different hygroscopic behaviour since species like sea salt or mineral dust
are predominately found in the larger size ranges. Nevertheless, this
assumption is deemed satisfactory since small particles have a minor impact
on light scattering compared to the effect of the larger sizes. In order to
obtain an ambient GF (GF for RHambient) the results at
RH = 95 % were recalculated for RHambient using
the semi-empirical κ-Köhler theory introduced by :
κ=GF(RH)3-1⋅1-awaw,
where aw is the water activity which can be inferred from the RH
and equilibrium droplet diameter (Dwet):
aw=RHexp4σs/aMwRTρwDwet.
Here, σs/a is the surface tension of the solution and/or air
interface, Mw the molecular mass of water, R the ideal gas
constant, T the absolute temperature and ρw the density of
water. We assume κ to be the same at all RH although this might
introduce some bias as former studies found changes of κ with RH at
elevated organic fractions (e.g. ). However, the
potential deviation due to this simplification is small in our case since the
GF is small anyway at the moderate RH encountered in this study.
The recalculated GF for RHambient were further used to
retrieve humidified aerosol particle size distributions from the measured dry
size distributions. The water uptake has also an influence on the index of
refraction which is taken into account for the Mie calculations by applying
a volume weighting mixing rule to determine the index of refraction of the
grown particles. Finally a humid scattering coefficient (μs,wet)
was calculated according to Eq. (). By propagating the
uncertainties of the single parameters in Eq. (), a mean
uncertainty of approximately ±18 % was found for
μs,wet. The ratio of μs,wet to μs,dry was
finally used to calculate f(RH) according to Eq. (). The
uncertainty in f(RH) amounted on average to ±25 %. Also
the ground-based data set was corrected for changes due to elevated RH by
using hygroscopicity results from the HTDMA (see Sect. for
more details). In this case the size distributions from the SPC-SMPS were
recalculated including adjustments for the RHambient in SPC.
Then the f(RH) was obtained with Eq. () using the dry
and humidified size distributions measured and retrieved in SPC and finally
it was applied to the directly measured scattering coefficients obtained from
the nephelometer to get μs,wet. The uncertainty in
f(RH)-SPC amounted on average to ±26 %, while
approximately ±27 % is found for μs,wet-SPC.
Aerosol absorption coefficient
A portable aethalometer (AE42, MAGEE Scientific; Berkeley, USA) was mounted
in the Zeppelin NT for a continuous measurement of the aerosol light
absorbing properties at seven wavelengths. This instrument monitors the
attenuation of light through a quartz fiber filter. The signal was then
corrected as proposed by for multiple scattering in
the filter matrix (“C value”) and the so called shadowing effect
(“f value”). A C value of 4.75 for λ = 520 nm was used
according to personal communications with J. P. Putaud who performed
a comparison between an aethalometer (model AE31) and a MAAP in summer 2012
in Ispra, Italy. The f value amounted to 1.06 on average. For the flights
a maximal attenuation of 70 % and a flow rate of 4 L min-1 were chosen.
The estimated uncertainty for this data set is ±20 %.
A multi-angle absorption photometer (MAAP; Thermo Scientific Carusso; Model 5012;
) was employed at the SPC ground station. It
measures the light attenuation and scattering by particles deposited on
a filter. The nominal wavelength is 670 nm, however, the actual
wavelength was found to be 637 nm .
A ±10 % uncertainty is estimated for these results. In order to
combine these measurements with those of the scattering coefficients, the
values were extrapolated to a wavelength of 520 nm using the
Ångström exponent (αa) from the multiple wavelength
measurement of the aethalometer. No aethalometer was available in SPC and
therefore αa was applied as obtained from the airborne data
set. To calculate αa from the airborne data set Eq. (1b)
from was applied, choosing the adjoining wavelengths λ1
and λ2:
αaλ1,λ2=-lnαλ1-lnαλ2lnλ1-lnλ2.
During this flight the absorption Ångström exponent αa
amounted on average to 0.93 ± 0.15 (mean ± SD).
Then Eq. () was applied to recalculate μa
measured by the MAAP to the wavelength of interest (520 nm):
μa,MAAP(520nm)=μa,MAAP(637nm)⋅520nm637nm-αa.
This introduces an additional uncertainty of 3 %, leading to a final
uncertainty in μa,MAAP (520 nm) of ±13 %. At
both locations μa was assumed not to vary substantially with
ambient RH. This assumption is justified by several reasons: μa
can potentially be enhanced by a shell around an absorbing particle
(“lensing effect”; ), however the magnitude of this
effect is not clear yet due to controversial findings
(e.g. for ambient aerosol). presented
theoretical calculations to investigate changes in the absorption coefficient
due to hygroscopic growth. In order to study RH effects, they compared the
dry and humidified responses and found only small enhancement of the
absorption coefficients at GF similar to the ones that were found in our
study. Therefore the effect is expected to be small for this case study.
Aerosol extinction coefficient
The extinction coefficient (μe) can be calculated as the sum of
the absorption and scattering coefficient.
μe(RH)=μa+μs(RH)
For the airborne as well as ground-based measurements the dry (RH < 30–40 %
as recommended by ) and wet (ambient RH)
extinction coefficients were retrieved. In this respect, the calculation of
the airborne scattering coefficient relies on the measured particle size
distribution, the retrieved index of refraction of the dry particles and
their hygroscopic growth. The most crucial parameter is the selection of the
index of refraction, which leads to the largest uncertainties of the
scattering coefficient. The absorption coefficient, on the other hand, is
assumed not to vary substantially with ambient RH and therefore no ambient
correction was applied. However, the ground-based absorption coefficient had
to be recalculated for a different wavelength using Eq. (). In
order to do so, the Ångström exponent obtained from the airborne
data set was used. The scattering coefficient is strongly dependent on RH and
was therefore corrected by measurements of the particles' hygroscopic growth
(see Sect. ). The propagated measurement uncertainties for
μe,dry and μe,wet amount to ∼ 6 and ∼ 24 %,
respectively, at the SPC ground site, while an uncertainty of
approximately 18 % is found for both on the aircraft. Please note that
for the airborne data set the relatively large uncertainty in f(RH)
is not propagated into the extinction results since the directly calculated
μs,wet are used for Eq. (). The ambient ground-based
results, however, are influenced by this uncertainty because f(RH)
is applied on the directly measured dry scattering coefficients from SPC.
In addition to the instrumentation described so far, a single wavelength
polarization diversity elastic lidar system was deployed at the SPC ground
station. This instrument uses a 532 nm pulsed Nd-YAG laser source,
with a pulse duration of 1 ns, energy of 400 µJ and
a repetition rate of 1 kHz. The lidar system collects the radiation
elastically backscattered from the atmosphere (Rayleigh scattering) by
separately detecting its parallel and cross polarization components with
respect to the polarization of the laser. Additional technical details of the
systems are presented in . The overlap of the laser beam
within the field of view (FOV) of the detector begins at a few tens of metres
from the system, and is complete at a few hundred metres (50 and
300 m, respectively, in the simple approximation of a conical laser
beam and telescope FOV). A nitrogen Raman scattering channel at
608 nm is also present, however these data are available only for
nighttime conditions. This channel, which collects a signal proportional to
the atmospheric molecular density, is used for the correction of the Rayleigh
signal coming from the region of partial superposition between laser and FOV,
the Partial Overlap Region (POR) where the backscattered signal is partially
lost. This correction is done by comparing the Raman signal received from the
POR with the molecular density profile obtained by collocated simultaneous
pressure and temperature balloon measurements, and thus retrieving an overlap
correction function, from the ratio of the Raman signal to the molecular
density. Uncertainties in the determination of the overlap function arise
mainly from the pressure and temperature uncertainties, and from the Raman
signal counting statistics and are reflected in inaccuracies in the
reconstructed signal of around 10 % at 100 m, rapidly
decreasing upward. The system provides a profile of backscatter ratio (R)
and volume depolarization ratio (DR) every 5 min for an elevation of up to
15 km, where R and DR are defined as
R=βa+βmβmDR=βa+βmcβa+βmp.
Herein βa and βm are the aerosol and
molecular backscattering coefficient, respectively, and the superscripts “p”
and “c” refer to their contribution in the parallel and cross polarized
backscattering. R and DR assume the value of 1 and 1.4 %
respectively, in regions supposed to be free of aerosol at a normalization
altitude z0, usually above 7 km. This normalization procedure
introduces an additional possible inaccuracy in the data produced, as the
derived backscatter and extinction coefficients at any height z below
z0, depends not only on the signal at z, but also on the extinction
between z and z0, on the ratio of the signal at z to the one at z0
and on the assumed values of R and DR at z0 . For the
data presented here, we performed a sensitivity test by varying the
normalization height z0 subjectively chosen in the region we assumed to be
free of aerosol. This resulted in a dispersion of the extinction data at
altitudes below 1000 m in the order of 15 %. Therefore, the
overall uncertainty for the lidar profiles is estimated to be approximately
25 %. Increased values of R indicate the presence of aerosol, while
departures of DR from its molecular value are indicative of depolarizing
(DR > 1.4 %) or not-depolarizing (DR < 1.4 %)
aerosol. The uncertainty associated with the data from the lidar used in this
study is extensively discussed in . The minimum relative
uncertainty on R for a 60 s measurement is 3 %. This
corresponds to a minimum detectable βa of 0.05 × 10-6 m-1 sr-1, at a signal to noise ratio of
100 %. This threshold level is reached close to the upper edge of the
POR, at approximately 800 m, where the backscattered signal attains
its maximum and decreases upwards because of increased distance from the
lidar, and downwards due to an increased loss of signal in the progressively
incomplete overlap between the laser beam and the telescope FOV. The
backscattered Rayleigh signal, which is only partially collected from the
POR, is multiplied by the overlap correction function to reconstruct its
entirety over that region. This correction is accepted if the reconstructed
signal exceeds the raw signal by no more than a factor of 20. This
corresponds to an acceptable reconstruction from approximately 100 m
upward. As already stated, this causes a possible inaccuracy in the order of
10 % of the reconstructed signal. Random errors, mainly arising from
poor signal statistics, add to this uncertainty and only these are reported
as error bars in our plots.
The inversion of the lidar signal is accomplished with the Klett method
using piecewise constant extinction to backscatter ratio
(a.k.a. lidar ratio LR) values:
LR=μeβa
and calibrating the profile by finding an atmospheric region supposed to be
free of aerosol particles, usually above 8 km. The value of LR
determines the aerosol extinction coefficient, once the aerosol backscatter,
βa, has been retrieved from the lidar measurements. The
values of LR to be used in the inversion are iteratively defined during the
inversion procedure itself, by inspecting at each step of the signal
extinction correction, the tentative values of R and DR. In regions of
different aerosol occurrence, desert dust is characterized by LR ∼ 50 sr
and DR greater than 10 % while
biomass burning aerosol commonly has a LR of around 60–70 sr and
a DR often lower than 10 % .
Table shows a list of LR used in our inversion, classified
according to R, DR and altitude. The Po Valley aerosol is predominantly of
continental origin and therefore LR values between 30 and 70 sr seem
to fit best as found using the CALIPSO model by for clean
and polluted continental aerosol particles, respectively, both at
λ = 532 nm. These values agree well with model results for
continental aerosol presented by who found LR values of
60 sr. Measurements performed in Southern Italy found LR values of
approximately 50 sr below 2 km at a wavelength of
351 nm . Discrepancies in LR may
arise from the selected method to retrieve it and from the exact location.
For instance, showed that comparing elastic lidar against
AERONET sun photometers generally yields higher LR compared to LR directly
from Raman lidars. In order to evaluate the LR assumption from literature
data, we performed Mie calculations for the backscatter coefficients using
the airborne in situ data. By applying Eq. () the LR was
calculated and yielded values between 51 and 67 sr, with a mean value of
58 ± 4 sr. These results strongly support our LR selection from
the literature. However, it is important to note that these in situ derived
LRs solely serve as plausibility check and were not used for any further
calculations or retrievals. We also compared aerosol optical depth (AOD)
obtained from the column-integrated lidar extinction (at 532 nm) to
the AOD from a sun sun-photometer (at 500 nm) at the same site
within the framework of the SKYrad NETwork
(http://atmos2.cr.chiba-u.jp/skynet/). The comparison of the AOD
variability during the time frame of the PEGASOS campaign showed good
agreement between the two data sets. For this period lidar derived AOD, using
a LR equal to 70 sr, yielded on average 7 % higher values than
those from the sun photometer. A sensitivity study changing the value of LR
to 50 and 30 sr resulted in underestimations of 5 and 25 %,
respectively. Thus, in this range of LR values, lidar agrees with the
sun-photometer in a column-integrated sense, within the reported limit. The
Supplement provides an in-depth discussion of lidar data treatment.
List of lidar ratios (LR) selected as a function of backscatter
ratio (R) and depolarization ratio (DR). For polluted continental air
masses, different values were tentatively employed for a sensitivity
study.
LR [sr]
R
DR [%]
Aerosol type
70 (30–50)
> 1.05; < 10
< 15
polluted continental
(clean continental)
50
> 1.05; < 10
> 10
Saharan dust
60–70
> 1.05; < 10
< 15
biomass burning