Introduction
The uptake of water by atmospheric aerosol particles depends on the
particle's hygroscopicity and the ambient relative humidity (RH). The
exchange of water vapour with the environment causes a change in size and
refractive index (RI) of aerosol particles and therefore directly influences its
optical properties. Especially the particle light scattering coefficient
σsp is strongly dependent on RH. The main quantity
describing this effect is called the scattering enhancement factor
f(RH,λ), which is defined as σsp(λ) at
elevated RH divided by its dry value
f(RH,λ)=σsp(RH,λ)σsp(RHdry,λ),
where λ denotes the wavelength, which will be omitted from now on for
simplicity. Nevertheless, one should keep in mind that all optical properties
are dependent on λ.
Long-term in situ measurements of aerosol scattering coefficients are usually
performed at dry conditions for example, recommends a RH below
30–40 %, but these in situ measured values differ from the
ambient- and thus climate-relevant ones. Knowledge of this RH effect is
therefore important for the calculation of the direct aerosol radiative
forcing see e.g.. In addition, the RH effect is also
important for the validation of model parametrizations
or for the comparison and validation of remote sensing
to in situ measurements
e.g..
The magnitude of f(RH) mainly depends on the aerosol chemical composition
and size. Several studies have experimentally determined f(RH) for
different ambient aerosol types using humidified nephelometer systems
see e.g.and
Sect. . Arctic and marine
aerosols usually show the greatest values of f(RH) which decrease with
increasing anthropogenic influence e.g. f(85 %,
550 nm)≈2–3.5;.
Continental aerosols e.g. f(85 %,
550 nm)≈1.8–2.8; and urban
aerosols e.g. f(85 %,
550 nm)≈1.3–1.6;
are observed with intermediate values. Low values are usually seen for
biomass burning aerosol e.g. f(80 %,
550 nm)≈1.01–1.51; or for highly polluted air masses
e.g. f(80 %, 550 nm)≈1.07–2.35;. Low values have also been reported for mineral dust which can
be transported over long distances e.g. from the Sahara to the European
continent e.g. f(85 %,
550 nm)≈1.2–1.7;. In
boreal environments, the aerosol particles are typically less hygroscopic
due to a large
contribution of organics . So far, the f(RH) of particles
representative for boreal regions has not been characterized in great detail.
This is the topic of the current study where f(RH) is analyzed combining
highly time resolved and detailed aerosol micro-physical and chemical
measurements. The results are further used to extrapolate the ground-based in
situ measurements, which include the RH effect on the particle light
scattering, to the atmospheric column using airborne measurements of the
particle number concentration and size.
The motivation for this study is based on two research questions:
What is the magnitude of the scattering enhancement factor f(RH) in the boreal forest region of northern Europe?
Can an optical closure between ground-based in situ and remote sensing aerosol measurements be achieved?
Instrumental
Particle hygroscopicity measurements
A humidified nephelometer (WetNeph) was deployed to measure the effect of
water uptake on the particle light scattering coefficient. The instrument is
described in detail by ; therefore only a brief description
will be given here. The WetNeph consists of a specifically designed
single-stream humidification system, where the aerosol first enters
a humidifier (at a flow rate of 9.5 L min-1) and then a drier before
the particle light-scattering coefficients are measured by an integrating
nephelometer at three wavelengths (λ=450, 525, 635 nm). An LED-based nephelometer (Ecotech Pty Ltd., Aurora 3000) was used, which is less
affected by the heat of the lamp that could influence the RH inside the
nephelometer. The WetNeph was set to the humidograph mode, where the RH
inside the nephelometer is periodically cycled between 35 and 40 and
90–95 % (slightly depending on the temperature inside the
measurement container). One full humidograph cycle (hydration and
dehydration) took 3 h. This set-up allows to measure the upper and
lower branch of the aerosol hysteresis curve separately. Dry scattering
coefficients were measured in parallel with a second (reference) nephelometer
of the same type as the WetNeph with an average RH inside the nephelometer
cell of 27.5±5.5 % (mean ± standard deviation; SD). From
these data, Eq. () is then used to calculate f(RH) for each
nephelometer wavelength.
All scattering coefficients were corrected for the truncation error and
non-idealities of the light source by the scheme described in
. First, the nephelometers were calibrated using particle-free air and CO2 as a span gas. Then both nephelometers were run in
parallel, measuring the same aerosol at the same RH, to determine the
relative differences between the two instruments. Relative differences
between 5 and 12 % were found for the three wavelengths, which was
accounted for when calculating the intensive parameter f(RH). In addition,
measured humidograms of polydisperse ammonium sulphate particles measured at
the site were compared to model predictions using the size distributions
measured by a differential mobility particle sizer (DMPS) system (with a diameter range of 6 to 600 nm, see
below), theoretical growth factors of ammonium sulphate and Mie theory
. Good agreement was found; however, the modelled values of
f(RH) were 5–10 % above the measured values, which can be
attributed firstly to the presence of few large particles that were not
included in the model calculations (due to the size cut of the DMPS) and
would lead to a lower predicted f(RH) , secondly to
the RH sensor's uncertainty (1–2 % absolute difference,
Rotronic HygroClip) and finally to the losses in the WetNeph system itself
between 2.5 and 5 %,. The relative measurement
uncertainty of f(RH) as an upper and conservative estimate is 20 % at
RH =85 % . The WetNeph showed a good
agreement to a novel commercially available humidified nephelometer system
(aerosol conditioning system (ACS1000) by Ecotech Pty Ltd.) for certain
periods of the campaign. At 85 % RH the median f(RH) agreed within
6 % for 525 nm (M. Laborde, personal communication, April 2015).
The humidograms of f(RH) can be described by an empirical two-parameter fit
e.g.:
f(RH)=a1-RH-γ.
The parameter a in Eq. () is the intercept at
RH = 0 % while γ describes the magnitude of the measured
f(RH). In previous work , the upper and lower
branches were fitted separately to the humidograms to investigate the
existence of aerosol deliquescence (sudden transition from the solid to the
liquid state of the particles; usually caused by pure inorganic salts).
However, no deliquescence was observed at Hyytiälä due to the dominance
of organic substances.
While f(RH) represents the hygroscopic growth as an optical measure, one
can also describe the hygroscopic growth by the change in particle diameter.
The hygroscopic growth factor g(RH) is defined as the ratio of the particle
diameter at elevated RH to its dry diameter
g(RH)=Dp,wet(RH)Dp,dry.
g(RH) was determined using a hygroscopicity tandem differential mobility
analyzer (H-TDMA), which is part of a volatile hygroscopicity tandem
differential mobility analyzer system. Detailed information on the
system can be found in . Four dry mobility diameters were
selected (Dp,dry=30, 60, 100, 145 nm) and their
humidified size distribution was measured at RH =90±2 % by
a second differential mobility analyzer (DMA) and condensation particle
counter (CPC, TSI Inc., Model 3772) system. The H-TDMA was calibrated with
ammonium sulphate particles at 90 % RH before the ambient sampling.
Particle absorption measurements
A filter-based absorption photometer (aethalometer, Model AE-31, Maggee
Scientific) was used to measure equivalent black carbon (EBC) mass
concentrations . The aethalometer is
a multi-wavelength instrument that measures the particle light absorption
coefficient σap at seven wavelengths by recording the
attenuation of light through a filter where particles deposit. The instrument
then converts the subsequent increase in attenuation to EBC concentrations
using a mass absorption cross section of
14 625 nmm2g-1 λ-1. The instrument was measuring
behind a Digitel PM10 ambient humidity inlet with a flow rate of
30 L min-1. A site-specific correction factor of C=3.35 to correct
for multiple scattering within the filter was applied
. A more detailed description of the aethalometer
measurements at the site is given by .
Particle size distribution measurements
The particle number size distribution was determined at ground level using
a DMPS for the fine mode (electrical
mobility diameter, Dp<1µm) and an aerodynamic
particle sizer (APS) for the coarse mode (aerodynamic particle diameter
Dp>1µm). The Hyytiälä-DMPS is a twin DMPS
set-up. DMPS1 has a 10.9 cm long Vienna-type DMA followed by a CPC
(TSI Inc., Model 3025). The measurement range is 3 to 40 nm
(electrical mobility diameter) with a sheath flow rate of 20 L min-1
and an aerosol flow rate of 4 L min-1. DMPS2 has a 28 cm long
Vienna-type DMA, followed by a CPC (TSI Inc., Model 3772). The measurement
range of DMPS2 is between 20 and 1000 nm with a sheath air of
5 L min-1 and an aerosol flow rate of 1 L min-1. The sheath
flows of the twin DMPS are dried to RH < 40 %, and continuously
controlled with regulating valves and inline flow metres. The aerosol flow
is brought to charge balance using a 14C radioactive source and the
flows are monitored using pressure drop flow metres. One measurement cycle
takes about 10 min. The Hyytiälä-DMPS is regularly calibrated and
checked with standard polystyrene latex spheres particles, higher precision
flow metres and has also been successfully intercompared to the ACTRIS moving
standard in 2009 . In addition to the twin DMPS, the
APS (TSI Inc., Model 3321) measured the size distribution in the aerodynamic
diameter range between 520 nm and 20 µm. The aerosol is
aspirated through a straight sampling line (tube diameter 16 mm,
length 4 m) to the instrument to avoid particle losses. The inlet is
at a height of 6 m above the ground and consists of a total suspended
particle inlet (Digitel Inc.). The inlet is heated to 40 ∘C to
prevent condensation and to ensure that fog droplets are evaporated and the
RH is below 40 %.
Particle chemical composition measurements
The aerosol chemical composition was measured by an aerosol chemical
speciation monitor (ACSM, Aerodyne Research Ltd.) which is permanently
deployed at Hyytiälä since March 2012. The instrument is a lighter
version of the Aerodyne aerosol mass spectrometer
developed for monitoring purposes. The ACSM inlet line had a PM2.5
cyclone filter to stop dust and pollen contamination. The inlet line is dried
using a Nafion dryer, reducing sample RH below 30 %. At the entry to the
instrument itself the sample aerosol is concentrated into a beam by a
standard aerosol mass spectrometer aerodynamic lens with a cut size of approximately 600 nm. The measured mass
is assigned to five main chemical species: sulphates (SO4), nitrates
(NO3), ammonia (NH4), chlorides (Cl) and organics (Org). For a more
detailed description on the data processing, the reader is referred to the
studies by , while more technical details on the
ACSM can be found in .
Assuming internally and externally well-mixed aerosol, the molar
concentrations of inorganic ions can be assigned to typically observed
inorganic salts: ammonium sulphate ((NH4)2SO4), ammonium bisulphate
((NH4)HSO4) and ammonium nitrate (NH4NO3). Since the amount of
chlorides at Hyytiälä was negligibly low, ammonium chloride (NH4Cl)
was excluded from the calculations. It was assumed that ammonium ions first
pair with SO4 ions to form ammonium sulphate and/or bisulphate –
depending on the molar ratio of NH4 to SO4 – with the remaining amount
of NH4 being available to form ammonium nitrate. Leftover NO3 was
considered to originate from organic nitrates. The loadings of NH4 were
typically too low to fully neutralize all of the observed SO4 and NO3.
Occasionally, the SO4 was left unneutralized, in which case they were
considered to originate from sulfuric acid (H2SO4). It should be noted
that the above calculations are very sensitive to the assumption of
well-mixed aerosol and additionally fail to account for possible organic
salts (e.g. organonitrate and organosulphate compounds). As these assumptions
are extreme in an ambient aerosol situation, the estimate must be considered
only a rough first approximation. However, it does provide some quantitative
results which we can use to predict f(RH), as shown in
Sect. .
Submicron elemental carbon (EC) mass concentration was measured using
a semi-continuous organic carbon (OC)/EC analyzer (Sunset Technologies Inc.). The instrument
measures the mass concentrations of OC and EC with
a time resolution of approximately 3 h. The device utilized
a two-step thermal–optical method for the determination of OC and EC. More
details can be found in and .
The chemical mass fraction Fi was determined by dividing the
concentrations of the individual components derived from the ACSM and
EC / OC analyses by the sum of all (excluding the OC from the EC / OC
analysis, which is covered by the ACSM measurement). The organic mass
fraction Forg was determined by adding the EC part (which is
known to have a low hygroscopicity) of the EC / OC analysis to the
organic components of the ACSM. The mass fraction is representative for
sub-micron particles only due to the experimental restrictions.
Auxiliary in situ instruments
Within the monitoring network, an integrating nephelometer (TSI Inc., Model
3563) is used to measure σsp, dry at λ=450, 550 and
700 nm. The instrument is located in the aerosol cottage behind an
switching PM1 and PM10 inlet (RH inside nephelometer cell
6.5±3.5 %). Here, only the PM10 measurements of
σsp, dry are used to retrieve the complex refractive index
and to compare the measurements of the WetNeph reference nephelometer to it.
The scattering coefficients were corrected for nonidealities of the light
source and the truncation error by the correction scheme of Anderson and
Ogren (1998).
Meteorological parameters like temperature, wind speed and direction or RH
were continuously measured along a 124 m high tower.
Airborne measurements
Vertical profiles of the aerosol size distributions were measured using
a Cessna 172 F aircraft as a platform
. The total particle concentration was
measured using an ultrafine condensation particle counter (TSI Inc.,
Model 3776) with a diameter cut off size of 3 nm. A scanning mobility
particle sizer SMPS, with a small Hauke-type DMA and CPC
(TSI Inc., TSI 3010) was used to determine the particle number size
distribution (mobility diameter size range of 10–270 nm). For
the SMPS the inversion by was used, and the calibration
corrections and turbulent tube losses were taken into account. Other
instruments inside the cabin included the Li-Cor 840 gas analyzer measuring
H2O and CO2 concentrations and a pressure sensor. Ambient air
temperature was measured using a PT100 sensor. A GPS receiver recorded the
flight path. The sample air inlet was a downscaled version of the inlet
design used with University of Hawaii's DC-8 . It was
situated under the right wing out from the propeller flow. The sample air was
led inside the aircraft via a stainless steel tube of length 4.2 m
and diameter of 22 mm. The flow rate of the inlet tube was between
45 and 50 L min-1.
The flights were conducted with a slow airspeed of ≈130 kmh-1. The ascend or descent rate was around
2.5 ms-1. Most of the research flights (23/30) were conducted
above the area surrounding the SMEAR II station at Hyytiälä. The other
flights were performed around Jämijärvi airport located 80 km
west of Hyytiälä. The flight profiles usually contained several flight
paths of around 30 km with constant altitudes and additionally
a climb up to 3.2 km. The direction of the flight paths was chosen to
be perpendicular to the wind direction at ground. The measurements are
described in more detail by .
Columnar and vertical measurements of aerosol optical properties
Columnar aerosol optical properties were measured using a sun photometer
(SPM, CIMEL CE-318) which has been operated at Hyytiälä since
February 2008 . The instrument was installed on
a 18 m high tower above the canopy of the forest surrounding the
station and is part of the AERONET network . It measures
direct sun irradiance to obtain the aerosol optical depth (AOD) at different
wavelengths (λ=340, 380, 440, 500, 675, 870, 1020 and
1640 nm) and the Ångström exponent (see Eq.
below). The absolute uncertainty of the AOD for this instrument type was
estimated by to be ∼0.01 for the visible and
near-infrared and ∼0.02 for the ultraviolet region. Moreover, other
optical and microphysical properties of atmospheric aerosols are routinely
retrieved using an inversion scheme developed by . The
calibration is carried out yearly by comparison with reference instruments,
after which final corrections are made and the data are available as
quality-assured level 2.0 data . The level 2.0 data has
been used in the following analysis.
In addition to the SPM measurements, data from a seven-channel Raman lidar
PollyXT; was included in
the data analysis. The lidar is located in Kuopio (62∘44′17′′ N,
27∘32′33.5′′ E, 190 ma.s.l.) which is 200 km
east-north-east of Hyytiälä. It is operated by the Finnish Meteorological
Institute within the Finnish observation network
and is part of the European Aerosol Research lidar Network
EARLINET;. PollyXT provides vertical
profiles of particle backscatter coefficients at wavelengths of 355, 532 and
1064 nm and the particle extinction coefficient at 355 and
532 nm. The system also includes a depolarization (532 nm)
and a water-vapour (407 nm) channel. The vertical resolution of the
instrument is 30 m.
Trajectory calculations
Air mass back trajectories were calculated hourly with the air parcel
arriving at an altitude of 100 m above the site using the HYSPLIT
model . The trajectories were calculated on
the basis of the Global Data Assimilation System data set (GDAS,
http://ready.arl.noaa.gov/archives.php). Along each trajectory,
additional parameters such as mixing layer (ML) height, temperature, RH and
column precipitation was calculated by the model. Each trajectory had a time
length of 10 days.
Time series of the scattering enhancement factor f(RH) at
RH = 85 % and λ=525 nm (bullet points) and the
dry particle light scattering coefficient at λ=525 nm (solid
line) measured at Hyytiälä. The error bars give the 95 %
confidence interval. The arrow indicates the period of airborne
measurements.
Probability density function (PDF) of (a) the measured
f(RH = 85 %), (b) the fit-parameter γ
(magnitude of f(RH)) and (c) the fit-parameter a
(intercept). The different lines show the result for the three nephelometer
wavelengths.
The surface residence time of an air parcel was then calculated by adding the
travel time of each trajectory for the entire measurement period on
a 1∘×1∘ longitudinal and latitudinal grid. Only
periods when the air parcel was within the ML were considered.
To further differentiate between the continental and maritime influence the
parameter ψ is introduced:
ψ=∫tstarttendρ(t)⋅ϵ(t)dt,
where tstart denotes the start and tend the arrival
time of the trajectory. The factor ϵ(t) in Eq. () is
+1 if the air parcel traverses within the ML above land, while it is -1
when the parcel traverses within the ML above oceans. The factor ρ(t)
accounts for the removal of the particles with an estimated half-lifetime of
one week (assuming a quadratic decrease with time). Other removal mechanisms
(e.g. due to precipitation) are not taken into account. By this definition,
ψ has as outer boundaries -1 (air mass traversed only above oceans)
and +1 (air mass traversed only above land). We are aware that this is
a simplified way of classifying the air masses; however, it will be shown
that ψ sufficiently describes the maritime and continental influence for
our purposes.
Results
Section describes the results of in situ measurements
of f(RH). Its correlation and the proposed parametrization to the
particle's chemical composition are discussed in Sect.
and . The following Sect. explains
the extrapolation of the ground-based in situ measurements to the atmospheric
column and compares the result to routinely performed
SPM measurements. Different hypothesis are discussed in
Sect. that can impact the comparison.
Mean, standard deviation (SD) and percentile values (prctl.) of the
scattering enhancement factor f(RH), the magnitude γ and intercept
a of the fitted humidograms.
Mean
SD
90th prctl.
75th prctl.
Median
25th prctl.
10th prctl.
f(85 %, 450 nm)
1.53
0.24
1.90
1.64
1.47
1.35
1.26
f(85 %, 525 nm)
1.63
0.22
1.95
1.74
1.57
1.48
1.42
f(85 %, 635 nm)
1.79
0.27
2.17
1.94
1.71
1.59
1.51
γ(450 nm)
0.24
0.07
0.34
0.28
0.22
0.19
0.16
γ(525 nm)
0.25
0.07
0.35
0.29
0.24
0.20
0.17
γ(635 nm)
0.30
0.08
0.41
0.35
0.28
0.23
0.20
a(450 nm)
0.96
0.07
1.07
1.00
0.95
0.91
0.88
a(525 nm)
1.01
0.05
1.08
1.04
1.00
0.97
0.94
a(635 nm)
1.01
0.05
1.08
1.05
1.01
0.98
0.95
Mean, standard deviation (SD) and percentile values (prctl.) of the
particle light scattering coefficient (σsp,dry), the
particle light absorption coefficient (σap,dry), the single
scattering albedo (ω0), the Ångström scattering exponent
(αsp, determined by a fit) and the main aerosol chemical
components (ACSM and EC / OC analysis). All optical properties are given
at dry conditions and were calculated to the wavelength of the WetNeph
nephelometer. The values are given for the time period when the WetNeph was
in operation (see Fig. ).
Mean
SD
90th prctl.
75th prctl.
Median
25th prctl.
10th prctl.
σsp,dry(450nm) [Mm-1]
42.03
25.42
79.57
52.91
34.07
22.84
18.35
σsp,dry(525nm) [Mm-1]
32.90
19.75
61.18
40.84
26.61
18.15
14.60
σsp,dry(635nm) [Mm-1]
27.19
17.65
51.01
34.34
21.27
14.62
11.00
σap,dry(450nm) [Mm-1]
2.11
1.19
3.45
2.54
1.90
1.32
0.92
σap,dry(525nm) [Mm-1]
1.82
1.00
3.05
2.22
1.62
1.14
0.82
σap,dry(635nm) [Mm-1]
1.51
0.82
2.63
1.89
1.35
0.94
0.66
ωsp,dry(450nm) [-]
0.95
0.02
0.97
0.96
0.95
0.93
0.91
ωsp,dry(525nm) [-]
0.94
0.03
0.97
0.96
0.94
0.93
0.91
ωsp,dry(635nm) [–]
0.94
0.03
0.97
0.96
0.94
0.93
0.91
αsp[-]
1.30
0.23
1.60
1.44
1.31
1.18
1.00
Organic mass conc. [µgm-3]
4.57
2.63
9.03
6.30
3.78
2.59
1.71
NH4 mass conc. [µgm-3]
0.37
0.15
0.57
0.46
0.36
0.26
0.20
SO4 mass conc. [µgm-3]
0.85
0.37
1.29
1.04
0.83
0.55
0.40
NO3 mass conc. [µgm-3]
0.20
0.11
0.37
0.26
0.16
0.12
0.10
Cl mass conc. [µgm-3]
0.01
0.01
0.02
0.01
0.01
0.00
0.00
EC mass conc. [µgm-3]
0.13
0.07
0.20
0.14
0.11
0.09
0.07
Probability density function (PDF) of the measured
f(RH = 85 %, 550 nm) at Hyytiälä (orange line) in
comparison to results obtained at other European sites where the same
instrument had been deployed (see legend; data taken from Zieger et al.,
2013). The result for Hyytiälä was linearly interpolated to
550 nm wavelength. The left panel shows the location of the different
sites.
Influence of water uptake on the aerosol light scattering coefficient at Hyytiälä
The time series of f(RH) at RH = 85 % was calculated by
averaging the humidograms every 3 h (one full RH cycle) and applying
Eq. () to the measurements. The result is shown in
Fig. together with the corresponding dry scattering
coefficient σsp, dry for λ=525 nm.
f(RH = 85 %, 525 nm) shows little variation throughout
the summer months with a mean value of 1.63±0.22.
f(RH = 85 %, 525 nm) decreases with increasing dry
scattering coefficient σsp, dry, indicating an increased
presence of less hygroscopic particles at high σsp, dry. The
probability density function (PDF) of the measured
f(RH = 85 %) for all nephelometer wavelengths and the entire
campaign is shown in Fig. together with the PDF of the fit
parameters used in Eq. (). A small increase of
f(RH = 85 %) with increasing wavelength is observed, similar
to observations made at Melpitz, Germany . This effect can
be reproduced by calculating the optical properties using Mie theory with the
input of the measured size distribution and chemical composition of the
particles. The fit-parameters γ and a consequently show a low
variation with a mean and SD value of 0.25±0.07 and 1.01±0.05
respectively. The value of a≈1 indicates the absence of hysteresis
effects. The mean, SD and percentile values of f(RH = 85 %)
are given for all wavelengths in Table together with the
fit parameters (see Eq. ). To bring our measurement results
into a broader context, Table shows the average
values for the main aerosol optical parameters (all calculated to the
nephelometer wavelengths) and the chemical composition measurements.
The f(RH) observed at Hyytiälä is remarkably low compared to other
sites. Figure shows the PDF at
RH = 85 % and λ=550 nm (linearly interpolated)
in comparison to other European sites where the same instrument with a
different nephelometer was used . High values of
f(RH) were measured for pristine maritime and Arctic aerosol found at
Ny-Ålesund, Spitsbergen (campaign mean and SD: f(85 %,
550 nm)=3.24±0.63), or aerosol dominated by inorganic salts as
recorded in winter 2009 at Melpitz, Germany (f(85 %,
550 nm)=2.77±0.37). Intermediate values were usually measured for
continental and anthropogenic influenced aerosol at Cabauw, the Netherlands
(f(85 %, 550 nm)=2.38±0.38), or free tropospheric
aerosol at Jungfraujoch, Switzerland (f(85 %,
550 nm)=2.30±0.33). The f(RH) values given above are campaign
averages; however, each site had its characteristics for specific air mass
types like marine aerosol, anthropogenic-influenced
aerosol or desert dust. For example, Mace Head in
Ireland showed distinct differences in f(RH) depending on the wind
direction; if the air had a maritime origin generally higher values were
observed (f(85 %, 550 nm)=2.28±0.19) in contrast to
wind coming from the island or continent with influence of anthropogenic
emissions (f(85 %, 550 nm)=1.80±0.26). A separation of
different air mass types for the other sites are given in Table 2 in
.
Results of the trajectory analysis (10-day backward calculations of
air masses arriving at Hyytiälä, black cross, averaged on
a 1∘×1∘ grid). (a) Total surface residence
time, (b) scattering enhancement factor (at RH = 85 %
and 525 nm), (c) organic mass fraction, (d)
equivalent black carbon concentration. Only concurrent times are shown when
all instruments were operated in parallel.
The trajectory analysis reveals further insights to the source of f(RH) as
shown in Fig. . Only concurrent times when all main in situ
instruments (WetNeph, ACSM and aethalometer) were running in parallel were
used. Figure a reveals that the main catchment area of the
air arriving at Hyytiälä was southern Finland, Russia, the Baltic Sea,
parts of Scandinavia and continental Europe as well as the Atlantic and
Arctic oceans. The f(RH = 85 %, 525 nm), the organic
mass fraction and the EBC concentration were separately averaged on
a 1∘×1∘ grid when the air parcel of the trajectory was
within the ML for each grid point. It is hereby assumed that the property did
not change along the trajectory. It can be seen in Fig. b
that air masses with eastern and continental origin had generally a lower
f(RH), while air masses traversing over oceans or originating from the
Arctic were characterized with elevated values of f(RH), which can be
explained by the contribution of hygroscopic sea spray particles transported
to Hyytiälä. However, no distinct deliquescence was observed in contrast
to other sites like Melpitz (Germany), Cabauw (The Netherlands) or
Ny-AAlesund (Spitzbergen), which can be explained by the high contribution
of organic substances at Hyytiälä. Figure c shows the
organic mass fraction is clearly elevated for continental air masses,
while it decreased for air masses having a maritime origin.
Figure d shows the spatial distribution of the EBC as
measured by the aethalometer. A strong source of EBC around St. Petersburg in
Russia and generally elevated concentrations of air masses coming from the
continent can be seen. No weighting or removal was considered for this
analysis since mainly intensive parameters are shown. In addition, the
analysis is also influenced by shadowing effects when air masses from
different origin are averaged on the same grid point to one mean value. This
can be avoided by using the factor ψ introduced in
Eq. (), which reveals the potential maritime and continental
influence. Figure shows the average values of
f(RH = 85 %, 525 nm), the EBC concentration and the
organic mass fraction vs. ψ. It can be seen that the scattering
enhancement is generally higher for maritime air masses, while it clearly
decreases with increasing continental influence. As an opposite trend, the
organic mass fraction steadily increases with more continental influence. The
EBC values show no significant trend compared to f(RH) or the organic mass
fraction.
Box plots of the (a) scattering enhancement
f(RH = 85 %, 525 nm),
(b) equivalent black carbon concentration and (c) organic
mass fraction vs. the factor indicating the maritime and continental
influence (see Eq. ; as boundary: -1 would be a trajectory
only traversing over oceans, +1 would be an air mass only traversing over
land). The central red mark is the median, the edges of the box are the 25th
and 75th percentiles and the error bars show the extent to the most extreme
data points that are not considered as outliers, while the outliers are
plotted individually (red crosses). Only concurrent times are shown when all
instruments were operated in parallel. The number of points are given in
Panel (a).
The measured scattering enhancement factors have been compared to further
in situ measured aerosol and meteorological parameters. No clear and
significant dependency was found when compared to the single scattering albedo,
aerosol size distribution parameters (total number concentration and mean
size), wind direction or wind speed. An exception was the small inverse
correlation (R2 = 0.45) that was found for the scattering Ångström
exponent (only when using the 450 and 525 nm scattering coefficients) and
the total particle surface area. This can probably be explained by the fact
that an increased concentration of mainly smaller particles (increased
Ångström exponent) were also composed of more organic components (lower
hygroscopicity), which overall caused a decreased f(RH). This is also seen
in the trajectory analysis, which revealed that air masses from the east
showed generally a higher Ångström exponent similar to the organic mass
fraction Forg (see Fig. c).
Comparison to the chemical composition measurements
The reason for the low f(RH) at Hyytiälä can be explained by the
dominance of organic substances in the particle's chemical composition, which
leads to lower particle hygroscopicity. As an example, the fit-parameter
γ (Eq. ) at λ=525 nm is plotted in
Fig. as a function of the organic mass fraction
Forg. The linear regression shows a clear anti-correlation
(squared Pearson's correlation coefficient: R2=0.77) with a decrease in
γ with increasing Forg i.e.
γ(525nm)=(-0.71±0.15)⋅Forg+(0.76±0.11);
retrieved from a weighted bivariate fit according totaking the SD of the
average values as an input for the uncertainty calculation. The
dominance of the organic mass fraction (mean ± SD: 0.7±0.11)
clearly determines the low values of γ and thus the low f(RH)
observed at Hyytiälä. For comparison, the values measured at Melpitz,
Germany, are added to Fig. for more details
see. The organic mass fraction at Melpitz of submicrometer
particles was substantially lower than at Hyytiälä (mean ± SD:
0.23±0.10). Although the γ values for Melpitz were measured at
a different time of year (winter) and showed a higher variability
(R2=0.50), they almost line up linearly with the observations made at
Hyytiälä. Due to measurement restrictions the total mass at Melpitz was
only differentiated between black and organic carbon, while the total mass at
Hyytiälä is determined from the elemental carbon of the EC / OC
analysis (organic carbon is assumed to be included in the ACSM organic mass
fraction). The ammonia mass fractions at Hyytiälä and Melpitz are also
linearly correlated, while the sulphate mass fraction did not show a joint
linear behaviour with the Melpitz data. The reason is that the aerosol found
at Melpitz during the winter months also contained large amounts of nitrate
which mainly formed ammonium nitrate (with a higher hygroscopicity than
organic aerosol), while the nitrate contribution at Hyytiälä was very
small and the sulphate mainly formed ammonium sulphate or ammonium
bisulphate,
which together with the organic contribution lead to a generally lower
hygroscopicity.
The fit-parameter γ (for λ=525 nm) vs. the
organic mass fraction Forg measured at Hyytiälä (green
bullets) and Melpitz, Germany (grey squares). The solid and dashed lines
represent the corresponding bivariate weighted linear
regressions.
A simplified parametrization for f(RH)
A summary of the linear fit parameters of γ vs. the chemical mass
fractions is shown in Table for the components which showed a
clear linear behaviour. The inorganic mass fractions, mainly sulphate and
ammonia, are clearly positively correlated with γ and f(RH), in
contrast to the anti-correlated organic mass fraction. This allows
the use of continuously performed chemical composition measurements at
Hyytiälä to predict f(RH) whether a humidified nephelometer is operated. It
can be done by taking the total organic or inorganic mass fraction as a proxy
for f(RH) and using the linear regression parameters given in
Table to calculate γ for each wavelengths. f(RH) then
follows by using Eq. (), assuming an intercept of a=1. The
variance of the intercept a can be used to estimate an uncertainty of the
f(RH) prediction (see Table ).
Parameters retrieved from a linear regression of the different
chemical mass fractions Fi (ACSM and EC / OC) vs. γ (fit
parameter for f(RH)) for the different nephelometer wavelengths. The
calculated uncertainty of slope and intercept of the used bivariate weighted
fit are given in parenthesis. The parameters for
NO3, Cl and EC are not given due to the low
correlation. The lower part (marked by an asterisk) shows the linear
regression parameters calculated in the same manner for the joint data set of
Hyytiälä (this study) and Melpitz for the components
which showed a joint linear behaviour. These values can be used to predict
f(RH) by using Eq. () (assuming an intercept of a=1).
Fi
Slope
Intercept
R2
450 nm
525 nm
635 nm
450 nm
525 nm
635 nm
450 nm
525 nm
635 nm
Organic
-0.70 (0.14)
-0.71 (0.15)
-0.79 (0.18)
0.74 (0.10)
0.76 (0.11)
0.85 (0.13)
0.79
0.77
0.79
NH4
3.19 (2.27)
3.37 (2.55)
3.56 (2.76)
0.03 (0.14)
0.03 (0.16)
0.05 (0.18)
0.50
0.47
0.58
SO4
1.00 (0.16)
0.99 (0.16)
1.15 (0.21)
0.08 (0.02)
0.09 (0.02)
0.11 (0.03)
0.79
0.80
0.78
NO3
-
-
-
-
-
-
0.05
0.05
0.11
Cl
-
-
-
-
-
-
0.00
0.00
0.00
EC
-
-
-
-
-
-
0.01
0.00
0.03
Inorganic
0.69 (0.18)
0.79 (0.18)
0.79 (0.23)
0.06 (0.04)
0.05 (0.05)
0.08 (0.06)
0.80
0.79
0.79
Organic*
-0.64 (0.06)
-0.67 (0.07)
-0.69 (0.07)
0.70 (0.04)
0.73 (0.04)
0.77 (0.05)
0.90
0.90
0.87
NH4*
3.44 (1.44)
3.56 (1.50)
3.54 (1.51)
0.02 (0.11)
0.02 (0.11)
0.05 (0.12)
0.82
0.84
0.83
Inorganic*
0.76 (0.13)
0.78 (0.14)
0.83 (0.16)
0.04 (0.04)
0.05 (0.04)
0.07 (0.04)
0.87
0.88
0.87
Numerical parametrizations of f(RH) using chemical mass fractions are only
sparsely published. proposed a similar parametrization of
γ using the mass fraction of organic matter and sulphate
(γs=-0.6⋅F̃org+0.9 with
F̃org=Corg/(Corg+CSO4) and
γs=ln(f(RH))/ln(1-RHref)/(1-RH), which is
similar to Eq. () if a=1; RHref denotes the dry
reference RH). This parametrization is limited to aerosol dominated by the
accumulation mode and is only given for λ=550 nm (P. Quinn, personal
communication, May 2015). Our results if calculated in the same manner
as described in show the same decreasing trend of γs
(for example for Hyytiälä: γs=-0.79⋅F̃org+0.96 and Melpitz: γs=-0.35⋅F̃org+0.81 at λ=525 nm). However, both data sets do
not show the same joint linear trend anymore because the organic mass
fraction of the parametrization by is calculated using the
organic and sulphate concentrations only. The aerosol at Melpitz, however,
had a significant contribution of nitrate, ammonia and black carbon which
needs to be included in the parametrization to retrieve a reliable estimate
on f(RH). In a more recent study, parametrized their
measurements of f(RH) from the Yangtze River Delta region in China in a
similar way as but adding also nitrate to the organic mass
fraction. A linear relationship of γs=-0.42⋅F̃org+0.54 with
F̃org=Corg/(Corg+CSO4+CNO3)
was found, which compares better to our results; however, the ammonia and
black carbon components are still missing in the linear relationship
presented by .
Table also states the linear regression parameters for the
joint Hyytiälä and Melpitz data sets. As mentioned above, the organic and
the total inorganic mass fractions showed a common linear behaviour and thus
a more general rule to predict f(RH) from aerosol chemical composition
measurements can be derived. Individual inorganic components like sulphate or
nitrate may show different functional dependencies individually for each
site; however, as the comparison to and
showed, it is important to include all major chemical constituents when
deriving a general parametrization of γ or f(RH) as has been done
here. Our parametrization for Hyytiälä is strictly spoken only
valid for the summer months when the fine mode is clearly dominated by less
hygroscopic organic substances. Verification during other seasons and adding
other sites is needed to allow a generalization of these findings. The
addition of the Melpitz findings from should only be seen
as a first step. Additionally, the parametrization may not be valid during
periods with substantially different coarse mode contribution which can have
a potentially large impact on the total f(RH) .
Extrapolation to the atmospheric column using aircraft measurements
The in situ measurements were extrapolated to the atmospheric column using
regular airborne profile measurements that were performed during the second
half of May until mid of June 2013. In total 17 profiles with collocated
cloud-free SPM measurements on the ground were available. The measurements
were binned in 200 m wide height levels (starting at 200 m a.s.l.).
The profile flights time took on average 2.5 h and included up to
three full ascends and descends. A comparison of the aircraft measurement at
the lowest flight level (200–400 m a.s.l.) to the ground-based
CPC shows a good agreement (R2=0.80, linear regression:
NtotCessna=1.17Ntotground-142 cm-3)
and slightly less particles by the ground CPC.
The AOD is defined as the vertical integral of the particle light extinction
coefficient σep:
AODλ=∫h0h1σepλ,hdh,
where h0 is the surface altitude, h1 is usually the top of the
atmosphere (e.g. when measured by a SPM) and
λ the wavelength. Here, h1 is the height of the highest profile
point reached by the aircraft.
To obtain the AOD from the in situ measurements the dry ground-based measured
σsp,dry (nephelometer) and σap,dry
(aethalometer) were first transformed to the respective SPM wavelength using
the Ångström law:
σspλ=kλ-αsp,
where k is the turbidity coefficient and αsp the
scattering Ångström exponent. Equation () can be
formulated for σap,dry in an analogous way. The sum of
σsp,dry and σap,dry yields the particle
light extinction coefficient σep,dry. We have limited the
extrapolation to SPM wavelengths that are close to the nephelometer
wavelengths to reduce the involved uncertainties. The in situ AODs are
therefore only calculated between 440 and 870 nm. To calculate
σep,dryλ,h at different altitudes the
total particle number concentration Ntot as measured by the
airborne CPC was used as a scaling factor ch. The in situ AOD
for the dry case then calculates as follows
AODdryin situλ=∫h0h1chσep,
drygroundλdh,withch=Ntot(h)Ntot(h0).
For the ambient in situ AOD, the particle hygroscopic growth at RH of the
different altitudes was now taken into account by using the ground-based
measured f(RH) and assuming that it does only depend on RH. This assumption
means that the particle chemical composition and intensive size distribution
parameter do not change with altitude. Eq. () then changes to
AODamb.in situλ=∫h0h1chf(RH,λ)σsp,
drygroundλ+σap,
drygroundλdh.
Note that the absorption coefficient is assumed not to change with RH. This
is a reasonable assumption at Hyytiälä due to the fact that the
scattering enhancement exceeds the absorption enhancement
and, even more importantly, due to the dominance of the
light scattering (i.e. campaign average for the single scattering albedo
ω0=0.94±0.03 at λ=525 nm, see
Table ), which in total will only induce a small
error. The f(RH) is linearly inter- or extrapolated to the SPM wavelengths.
To test the influence of the layer above the maximum flight altitude an
exponential decrease of the total number concentration was assumed (with
c(h)=c(hi)exp(-0.25h) above the maximum flight altitude, where c(hi)
is the scaling factor of the last height bin and h the altitude up to
7 km). This is a reasonable assumption only for cases without clear elevated
layers, which was most likely only given for the first half of the airborne
observation period (see Sect. ). The in situ AOD with
the exponential decreasing profile above the maximum flight level is only
calculated for the dry case since no RH measurements are available above the
maximum flight altitude.
Example of the ground-based in situ measurements extrapolated to the
atmospheric column. Particle light extinction coefficient
σep (at λ=500 nm) measured at the surface at
ambient RH (along the tower at 17, 67 and 124 m; green bullets),
surface extinction coefficient weighted with the relative changes in the
total number concentration measured by the aircraft CPC (dry, orange points)
and at ambient conditions (violet points) with the RH measured on board the
aircraft (blue points, upper axis). The red points are dry values of
σep above the maximum flight altitude assuming an
exponential decreasing particle concentration. The error bars denote the 25th
and 75th percentile values. (a) Result for the 23 May 2013.
(b) Result for the 02 June 2013.
(a) Time series of the maximum altitude during the aircraft
profiling. (b) Time series of the AOD at λ=500 nm
measured by the sun photometer (SPM, grey curve), determined from the
ground-based dry extinction coefficient and the airborne CPC as scaling
factor (orange curve) and determined from the ground-based extinction
coefficient at ambient conditions (violet curve). The red dashed curve
represents the in situ derived AOD when an exponential decreasing profile is
assumed above the maximum flight altitude (at dry conditions). The error bars
denote the distance to the 25th and 75th percentile values, while the centre
point gives the median value for each profile. (c) Relative
difference of in situ derived AOD compared to the SPM measurement
((AODin-situ- AODSPM)/ AODSPM⋅100% at
λ=500 nm). (d) Relative difference of dry in situ derived
AOD compared to the SPM measurement for different SPM wavelengths.
To calculate the in situ AOD the atmosphere above was separated into 200 m
wide levels in which the CPC measurements were averaged to determine c(h)
for each layer starting at 200 m a.s.l. (close to the top of the canopy and
location of the SPM). Two example profiles showing
the in situ derived profiles are presented in Fig. .
For comparison, the ambient extinction coefficient measured at the ground is
shown together with the RH profile. As a test for the variability, the
calculations were repeated by using the 25th and 75th percentiles as lower
and upper boundary respectively. In the first example, the top of the ML is
clearly seen at around 1500 m. The particle light extinction
coefficient sharply decreases above the ML. The RH effect is significant but
not very strong due to the low hygroscopicity of the organic-dominated
aerosol at Hyytiälä and the low RH profile during that time of day (RH
varied within the ML between 50 and 70 % while it decreased to
20 % above 2000 m). Integrating the ambient extinction
coefficient profile yields an AODamb.in situ of 0.018
at λ=500 nm, while the SPM measured a value of 0.055. The
second profile example (Fig. b) shows the result for
the 02 June, where no clear ML transition can be observed. The extinction
coefficient still is elevated even at the maximum flight level of
2700 m. An integration of the ambient extinction coefficient profile
gives an AODamb.in situ of 0.1 at
λ=500 nm, while the SPM measured
a value of 0.37. The time series for all profile retrieved AOD values and the
SPM measured ones are depicted in Fig. for
λ=500 nm together with the maximum flight altitude. The in situ
derived values follow the course in time of the direct AOD values of the
SPM. However, they are 2–3 times smaller than the
directly obtained ones
(Fig. c).
Figure b and c also reveals
that the addition of an assumed exponential decreasing profile above
approx. 3 km only marginally leads to an increase of the in situ derived dry
AOD. This points towards the fact that most of the particles were captured by
the aircraft profiling, if the assumption of an exponential decrease in
particle number concentration is valid. However, this assumption is most
likely not valid for the second half of the aircraft profiling period. As can
be seen in Fig. b, the AOD increases in the beginning
of June due to long-range transport of mineral dust in elevated layers (see
Sect. ). The WetNeph was not in continuous operation
between 08 and 15 June 2013 due to computer failures and thus the ambient
AODamb.in situ was not retrieved for this period.
The calculations were done for all SPM wavelengths between 440 and 870 nm
which are close to the spectral region of the nephelometer. Figure d shows that the relative difference of the dry
in situ derived AOD to the SPM measured values increases for larger
wavelengths. These differences are more pronounced for the period of
potential long-range transported mineral dust.
The following hypotheses are brought forward to explain the clear
disagreement between in situ derived and directly measured AOD:
assumptions made to calculate
AODin-situ;
inconsistencies within the in situ measurements;
missing coarse mode particles (Dp>1µm) and general sampling losses within the ground-based in situ
measurements;
removal by dry deposition within the canopy;
aerosol layers above the maximum flight altitude.
The hypotheses will be discussed in detail in the following section.
Discussion
Influence of general assumptions being made
The main assumptions that were made in Sect. can all have a
potential influence on the disagreement between in situ derived and measured
AOD values. The first main assumption is to use the total particle number
concentration as scaling factor c(h) in Eqs. () and
(). It should be noted here that the results are in a similar
range if the particle surface is being used to calculate c(h); however, that
factor would omit optically active particles above the upper size limit of
the airborne SMPS (see Fig. b) and therefore we
prefer to take the total concentration to determine c(h).
To calculate the ambient extinction, it was assumed in Eq. () that
the particle light absorption enhancement is negligible. As mentioned above,
this is justified for this site due to the low absorption enhancement effect
compared to the scattering effect and the overall dominance of particle light
scattering when determining the particle light extinction coefficient
.
For the ambient case, it was additionally assumed that the f(RH) is the
same within the column as measured at ground and therefore only depends on
the RH at different altitudes. This assumption implies that the chemical
composition (hygroscopicity) and mean size is constant throughout the
atmospheric column. This assumption is most likely fulfilled for a well-mixed
boundary layer; however, it will not be valid for lofted or separate layers
during episodes with long-range transported air masses. During the summer
months at Hyytiälä, however, the columnar RH was always moderate and low
in addition to the fact that particles are generally less hygroscopic at this
boreal site and, therefore, the overall effect of the constant
f(RH) assumption was probably small compared to the hypotheses discussed
below.
Consistency of in situ measurements: optical closure study
To prove the consistency of the optical and
microphysical aerosol in situ measurements, a closure study based on Mie
theory was performed. The particles were assumed to be
spherical, homogeneous and internally mixed. As input, the particle number
size distribution measured by the DMPS and APS was used (the APS and DMPS
size distributions were merged at the last DMPS size bin). The complex
refractive index was inverted from the dry scattering (nephelometer) and
absorption coefficient (aethalometer) measurements and the measured particle
number size distribution using Mie theory . Only the
measurements from the continuous aerosol monitoring program were used for the
retrieval since they were also located inside the aerosol cottage. The
calculation was done incorporating the TSI nephelometer illumination
sensitivity and the specific scattering angles to avoid the truncation error
. For λ=450 nm a mean value for the RI of
(1.56±0.07)+(0.008±0.005)i was calculated, while
(1.53±0.06)+(0.008±0.005)i and (1.50±0.07)+(0.008±0.005)i were
calculated for λ=550 nm and λ=700 nm
respectively. These retrieved real parts of the RI for Hyytiälä are close
to the values of ammonium sulphate e.g. 1.536+10-7i at
λ=450 nm;. The result of the Mie calculations is
shown in Fig. a, in which the relative differences between
prediction (Mie calculation) and measurement are shown for all nephelometer
wavelengths. The monitoring nephelometer (located in the cottage) is in
almost perfect agreement to the calculation which is reasonable since the
same measurement was used to retrieve the RI. However, the little variation
proves that it is justified to use an average and fixed RI for each
wavelength for the entire period. The calculated σsp,dry for
the dry nephelometer used within the WetNeph system (located in the campaign
containers) are clearly underestimated by the model calculations (on average
8–30 %, see Fig. a). This corresponds to general
differences between the dry monitoring nephelometer in the cottage measuring
less particle light scattering than the reference nephelometer of the WetNeph
located inside the container (see Fig. b). The lower measured
scattering coefficients of the cottage nephelometer are in correspondence to
the underestimation of the measured particle number size distribution, which
is an input to the Mie calculation. Therefore, particle number concentration
and light scattering measurements of the monitoring measurement inside the
cottage were affected by the same loss effect. Almost the same result is
obtained when the RI of ammonium sulphate is taken. Small parts of the
disagreement could come from general calibration issues of the nephelometers
used in the WetNeph set-up. The larger variation of the WetNeph reference
nephelometer (the error bars denote the 25th and 75th percentile values)
suggests that the container site experienced more variation in aerosol
concentration compared to the cottage site inside the forest.
(a) Result of the optical closure study. Relative
differences of the predicted to measured scattering coefficient (dry) and
scattering enhancement factor (at RH = 85 %) for the different
nephelometer wavelengths. The circle denotes the median value and the error
bars the 25th and 75th percentile values. (b) Comparison of the dry
nephelometer measurements (σsp,dry) between cottage
(monitoring) and container (WetNeph). The values of the cottage nephelometer
were interpolated using Eq. ().
The differences in the scattering coefficients cancel out when the scattering
enhancement is calculated. In a first test, the hygroscopic growths factors
g(RH) (Eq. ) of the HTDMA were taken details on the
f(RH) calculation can be found in. The g(RH) values were
interpolated between the measured dry diameters of Dp=30 and
145 nm. Above 145 nm, the values of g(RH) were assumed to
be the same as the one measured at Dp=145 nm (similar
for Dp=30 nm). The calculated values of f(RH) using
the HTDMA measurements lie on average within the range of the measured values
(Fig. a). A slight disagreement for the larger wavelengths (on
average 12 % at 635 nm) is found. As second test, the values
of g(RH) were calculated using the ACSM and EC / OC measurements. The
value for pure organics was first assumed to be
gorg(RH = 90 %) = 1.2
and secondly assumed to be
gorg(RH = 90 %) = 1.05, a value recently derived
for isoprene-dominated organics at Hyytiälä . The
calculated values using the original value of
gorg(RH = 90 %) = 1.2 are systematically higher
than the direct measurements (≈30%), while the lower value
of gorg(RH = 90 %) = 1.05 delivers an improved
agreement. This points towards the importance of the hygroscopic growth
factor, which is especially for low hygroscopic substances important when
calculating f(RH) see Fig. A1 in.
Summarizing the optical closure study, one can conclude that the different
in situ measurements provide consistent results. However, the differences
found in the scattering coefficients measured by the monitoring and reference
nephelometers point towards losses. Partitioning effects of semi-volatile
organics or nitrate components (due to the low
concentration to a lesser extent, see Table ) that
could have caused a potential decrease in the overall particle properties
cannot be ruled out completely. Although it is believed to have a minor
effect during the summer months and daytime in situ measurements at this
site. Smaller differences can additionally be explained by the simplified
assumptions taken for the Mie calculations (e.g. internal mixture,
homogeneous and spherical particles, no size dependence of the refractive
index, specific values for g(RH)).
(a) Average particle number size distribution measured at
ground and within the lowest flight level by the aircraft
(200–400 m a.s.l.). (b) Aerosol scattering size
distribution calculated using Mie theory for the wavelengths of
500 nm (RI = 1.51). The centre lines show the median, while the
corresponding shaded areas denote the 25th and 75th percentile values.
Particle losses
The SPM was placed on a tower above the forest canopy
(∼198 m a.s.l.), while the in situ measurements were performed on
ground below the canopy (∼180 m a.s.l.). Particles may have been lost
within the canopy by dry deposition before reaching the inlet
, which includes removal through
Brownian diffusion (mainly for fine mode particles below
Dp<100 nm) or through impaction or interception (mainly
for coarse mode particles above Dp>1000 nm).
performed aerosol flux measurements using the eddy
covariance technique at Hyytiälä and found that only 35 % of the
particles penetrated through the canopy at low wind speeds. At higher wind
speeds and correspondingly stronger turbulent conditions only 10 %
of all particles reached the ground. The study by was
performed in spring, while our measurements were done in summer months with
probably more turbulence and thus higher deposition losses. In addition,
particle losses could have also occurred within the inlet and tubing itself.
However, this is rather unlikely since the optical closure study has shown
the consistency of the optical and microphysical aerosol measurements.
Coarse mode fraction of the particle light scattering coefficient
vs. the sun photometer wavelengths. The black centre line shows the median
value, while the shaded area denote the 25th and 75th percentile value range
for the period with airborne measurements.
Figure a shows the average particle number size distribution
measured at the ground and by the aircraft within the lowest layer. For small
particles below 100 nm, the aircraft measured on average higher
concentrations (up to 40 %) than the ground-based instrument.
However, for the optically important size range above 100 nm, both
size distributions agree surprisingly well. Figure b depicts
the scattering size distribution calculated using the measured size
distributions and Mie theory. Here, both size distribution measurements agree
until the maximum diameter of the aircraft SMPS is reached. Unfortunately,
above Dp>270nm the aircraft did not record the size
distribution and thus missed information on the optically important part of
the aerosol size spectrum.
The relative disagreement between in situ derived and measured AOD values
increased for larger wavelengths (see Fig. d), which
points towards an influence of large particles which are not sufficiently
sampled by the in situ instruments. Figure shows
the calculated scattering coarse mode fraction (defined as the scattering
coefficient for particles above Dp>1µm divided by
the total scattering coefficient both calculated using Mie theory) for all
wavelengths used in the SPM measurements. The
calculation was done for all time periods with corresponding profiles using
the particle number size distribution measurement on the ground. With
increasing wavelength more light scattering will be due to coarse mode
particles. At λ=1020 nm, for example, it is already
50 % for the here measured aerosol. A few losses of supermicron
particles can therefore explain the observed differences.
Aerosol optical depth (AOD) of fine mode particles derived from
AERONET vs. the in situ derived value using Mie theory and the measured size
distribution (at 550 nm). The error bars denote the range of the 25th
and 75th percentile values, while the centre points mark the median value.
The AOD for the fine mode fraction (Dp<1µm) was
estimated by taking the measured particle number size distribution at ground
and applying Mie theory (taking the RI from the Mie inversion, see
Sect. ) which results in the extinction coefficient for
submicrometer particles only. The calculation of in situ AOD for the fine
mode fraction followed in the same manner as described above (using
Eq. ). The comparison of the derived values to the AERONET
inverted fine mode AOD is shown in Fig. . A high
correlation was found (R2=0.84) and a linear least-squares regression
revealed that the AERONET values were significantly higher (slope of 1.53)
compared to the in situ derived values. Again, this indicates that besides
the missing coarse mode also the loss of fine mode particles contributed to
the found disagreement. These particles could have been fine mode particles
above the maximum flight altitude (see Sect. ) or
particles possibly lost through dry deposition within the canopy.
Elevated layers
As discussed above, the airborne sampling was only performed to a maximum
altitude of 3.2 km. Thus, elevated layers contributing to the
columnar AOD could have been missed by the aircraft. The time series of the
AOD in Fig. b already showed an unusual increase of
the AOD (to 0.35 at λ=500 nm; starting approximately on the
02 June 2013) compared to the average values measured at Hyytiälä
0.12±0.04,. Figure shows lidar
profiles of the aerosol backscatter coefficient (at
λ=1064 nm) recorded at Kuopio during the airborne campaign.
While there are no significant elevated layers before 28 May, clear elevated
layers are seen above 3000 m from 30 May until 03 June. Air mass back
trajectories showed that the air originated from the Arabian peninsula and
thus could consist of layers of mineral dust particles. In addition, the
depolarization channel showed values indicative for non-spherical particles.
Figure also gives the percental contribution from the layer
above 3 km and below 7 km to the total AOD derived by the
lidar. For the period before 02 June, 15–25 % of the AOD was
attributed to aerosol particles in elevated layers, while elevated layers
contributed between 60 and 80 % to the AOD between 02 and 03 June
(assuming a constant lidar ratio). These percentages are in correspondence to
the relative differences calculated for the in situ derived AOD vs. the
measured values by the SPM during the period of long-range transported
mineral dust (see Fig. d). It should be mentioned that
the comparison to the lidar profiles measured at Kuopio is only of
qualitative nature to demonstrate the effect of lofted layers due to
long-range transport.
Conclusions
The effect of water uptake on the particle light scattering coefficient was
investigated at a boreal site using a humidified nephelometer system.
Compared to other major aerosol types the aerosol light scattering
enhancement factor f(RH) shows low values with little variation (at
RH = 85 % and λ=525 nm a mean value of
1.63±0.22). This is attributed to the dominance of the organic mass
within the submicron range, to which f(RH) clearly correlates
(R2=0.77–0.79). f(RH) can thus be estimated using the continuous
chemical composition measurements when no direct f(RH) measurement is
available for the aerosol found during summer months at Hyytiälä.
A trajectory analysis revealed that higher values of f(RH) and higher
inorganic mass fractions at Hyytiälä were due to hygroscopic sea spray
particles transported to the site.
Aerosol backscatter coefficient profiles measured by the PollyXT
lidar in Kuopio (200 km east-north-east of Hyytiälä) before
(a) and during (b) the long-range transport period
(λ=1064 nm). The percental numbers in parentheses denote the
contribution of elevated layers above 3 km to the total AOD of the
lidar profile.
The measurement of the f(RH) allowed to estimate the particle light
extinction coefficient at ambient RH. This was then used to extrapolate the
ground-based in situ parameters to the atmospheric column using aircraft
measurements of the particle number concentration as a scaling factor. The in
situ derived AOD were correlated to the sun
photometer measurements; however, a clear underestimation of the AOD by at
least a factor of 2 was found. To investigate the reasons for this
disagreement, different hypotheses were brought forward and discussed. An
optical closure study showed the validity of the ground-based in situ
measurements and showed that a lower hygroscopic growth of the organic
compounds resulted in an improved agreement with respect to f(RH). The
discrepancy of the in situ derived AOD increased for larger wavelengths,
pointing towards an underestimation of coarse mode particles which might have
been removed by the canopy or were not sufficiently sampled by the in situ
instruments. In addition, elevated layers observed by a lidar at Kuopio can
explain part of the found disagreement. The remaining differences are
speculated to come from dry deposition within the canopy.
This work demonstrated the difficulties faced when using ground-based in situ
measurements for the validation of remote sensing (e.g. sun photometer and
later even satellite) measurements. Consequently, more research work and
improved measurements are needed for integrating in situ measurements for the
validation or comparison of remote sensing retrievals. For sampling sites
located in forest environments, the removal by the canopy of fine and coarse
mode particles has to be included when analysing time series of aerosol
optical properties.