Introduction
Biomass burning aerosol (BBA) can act as efficient cloud condensation nuclei
(CCN) and form cloud droplets. Fires can therefore influence cloud formation,
growth, reflectance, precipitation and lifetime (Kaufman et al., 1998; Warner
and Twomey, 1967). The contribution of CCN from fires results in higher
concentrations of cloud droplets, which yield whiter clouds that generally
survive longer than clouds with fewer droplets (Platnick and Twomey, 1994).
While greenhouse gases and black carbon emitted from fires absorb radiation
and have a warming effect, the influence of solar radiation scattering by
organic material and the production of CCN has a cooling effect on the
Earth's lower atmosphere. The net forcing of carbonaceous combustion aerosol
is thought to have an overall global cooling effect (Spracklen et al., 2011;
Ward et al., 2012). The complexity arises from variability in emission
factors, BBA size, composition and aging. This contributes to a large
uncertainty that these fires have on the radiative budget (Carslaw et al.,
2010). Thus detailed measurements of the physical and chemical properties of
BBA from all regions in different seasons are essential in determining their
impact on clouds (Spracklen et al., 2011). Very few studies have taken place
within Australia, despite Australia contributing an estimated 15 % of
yearly global burned land area (van der Werf et al., 2006). Australian
studies have been typically focused on fires in the southern continent
(Lawson et al., 2015) or east coast cane fires (Warner and Twomey, 1967). The
extent to which dry season fires in northern Australia impact CCN concentrations
has not been explored in detail.
Most of central northern Australia is unpopulated and is characterized by
savannah vegetation with grasslands, shrubs and scattered eucalypt trees.
Although the most devastating fires burn in the densely populated southern
regions of Australia, the vast majority of the continent's fires occur in the
north and are responsible for more than half of land area affected by fires
(Russell-Smith et al., 2007). During the dry season (May until November)
thousands of fires burn via prescribed burning and natural or accidental
ignitions. The frequency and severity of these fires increases as the season
progresses from the early dry season to the late dry season (Andersen et al.,
2005). Under Aboriginal management, fires were lit in the late dry season in
order to prepare for the wet season. These late dry season fires may have
been lit intentionally to trigger the onset of rainfall following the
formation of pyro-cumulus clouds, among other ecological reasons (Bowman and
Vigilante, 2001; Bowman et al., 2007). Under non-Indigenous management, early
dry season prescribed burns are commonplace in order to reduce the severity
of late dry season fires (Andersen et al., 2005). Outside of the only major
urban centre in this region, Darwin, prescribed burns are the dominant source
of accumulation mode aerosol particles (Mallet et al., 2016). Although long-range mineral dust sourced from the central Australian desert was observed
during SAFIRED (Winton et al., 2016), the number concentrations of these
particles in the sub-200 nm size range was likely to be negligible. Thus
these prescribed burns will dictate CCN concentrations in the region.
The vegetation (fuel) type, burning conditions and atmospheric aging
determines the size, composition and the hygroscopicity of BBA, and in turn
their ability to act as CCN. BBA is typically a mixture of elemental carbon
(EC) and organic carbon (OC) and can contain inorganic material (Reid et al., 2005). The
precise OC composition of primary BBA can vary greatly depending
on the fuel type and these organic constituents can be weakly or highly
hygroscopic (Carrico et al., 2010; Mochida and Kawamura, 2004; Novakov and
Corrigan, 1996; Petters et al., 2009). The hygroscopicity of BBA can change
with oxidation and with the condensation or evaporation of volatile organic
compounds (VOCs) through atmospheric aging (Hennigan et al., 2011). Smog
chamber experiments have shown that after a few hours of simulated
photochemical aging, the hygroscopicity of BBA converges to weakly
hygroscopic for many different fuel types (Engelhart et al., 2012; Giordana
et al., 2013).
While laboratory-based measurements are useful in understanding the physical
and chemical processes that determine the hygroscopicity and composition of
aerosol, they do not necessarily represent ambient conditions. Due to
feasibility, however, direct ambient measurements of the CCN activity of
smoke plumes are rare (Lawson et al., 2015) and more measurements are useful
in assessing the validity of climate models. A previous preliminary study of
the CCN activity of savannah fires in the northern Australian early dry season
reported moderately hygroscopic BBA (Fedele, 2015), speculating that the
aerosol is mostly made up of aged biomass burning particles with a coating of
secondary organic aerosol. While these measurements took place over a short
period, there was a discernable slight increase in the hygroscopicity of BBA
during the day. Diurnal patterns in hygroscopicity have been observed in
boreal environments (Paramonov et al., 2013) and in the southeast United
States (Cerully et al., 2015), attributing increases in daytime
hygroscopicity to the photochemical oxidation of organic aerosol.
Some studies suggest that the impact of composition, and therefore changes in
BBA hygroscopicity due to photochemical aging, on CCN concentrations is much
lower than the impact from the aerosol size distribution (Dusek et al., 2006;
Petters et al., 2009; Spracklen et al., 2011). Under this assumption, changes
to the activation diameter resulting from a change in hygroscopicity are less
important than the size distribution of the BBA. Other studies have shown
that while this is true for moderately and strongly hygroscopic particles,
cloud droplet number concentrations are moderately sensitive to weakly
hygroscopic particles (Reutter et al., 2009; Sánchez Gácita et al., 2017).
The total number of detected hotspots (confidence that hotspot is
fire > 50 %) between 30 May and 1 July 2014
in Australia.
Smoke from biomass burning can be transported over intercontinental distances
and can reach the upper levels of the atmosphere (Andreae et al., 2001;
Dirksen et al., 2009). Aircraft measurements during the early and late dry
season in northern Australia, however, suggest that smoke from fires in this
region is contained within the planetary boundary layer (Ristovski et
al., 2010; Kondo et al., 2003). Trade winds collect and carry this smoke
northwest over northern Australia, the Timor Sea and the tropical warm pool.
Cloud albedo is more sensitive to aerosol concentrations in pristine
environments (Twomey, 1991). The biomass burning that occurs during the dry
season is the dominant source of particles in northern Australia, and is
therefore likely to influence aerosol–cloud interactions over the tropical
warm pool in the Timor Sea.
This paper presents a comprehensive data set of the particle size, chemical
composition, hygroscopicity and CCN properties of BBA generated from fires
in the dry season in this region. The impact of BBA size and hygroscopicity
on CCN activation is discussed in detail. These parameters will be useful
in climate models to assess the magnitude of climate forcing by BBA in
aerosol–cloud interactions.
Experimental design
Sampling took place at the Australian Tropical Atmospheric Research Station
(ATARS; 12∘14′56.6′′ S, 131∘02′40.8′′ E), Gunn Point, in the
Northern Territory of Australia as a part of the Savannah Fires in the Early
Dry season (SAFIRED) campaign (Mallet et al., 2016). The research station is
located near the tip of a small peninsula with close proximity to the Timor
Sea and Tiwi Islands. The territorial capital, Darwin, lies 20 km to the southwest of the station. Savannah vegetation with scarce human settlements
transitions over hundreds of kilometres to the south into the desert regions
of central Australia. Sampling for the SAFIRED campaign occurred in June 2014
at ATARS. This period is the early dry season in this region, where strategic
small-scale controlled burns are performed in order to reduce the frequency
and intensities of fires in the late dry season in October and November.
Despite sampling occurring during winter, daily temperatures can reach well
above 30 ∘C such that accidental and natural fires can also occur.
Throughout the sampling period, thousands of fires were observed in northern
Australia. This led to strong biomass burning signatures detected at the
station, with numerous instances of very intense BBA events from both distant
and close fires. A full overview of the campaign, including meteorological,
gaseous and aerosol measurements, is presented in Mallet et al. (2016).
Sentinel Hotspots, an Australian national bushfire monitoring system, was
used to investigate the number of daily fires in the region. Sentinel uses
data from the MODIS (Moderate-resolution Imaging Spectroradiometer) sensors
onboard the Terra and Aqua NASA satellites and the VIIRS (Visible Infrared
Imaging Radiometer Suite) sensor onboard the NASA/NOAA Suomi NPP satellite.
These satellites fly over northern Australia once per day between 11:00 and
15:00 local time. Although fire locations are therefore limited to those
that are burning during these times, Sentinel is still useful in providing
information on the spread and number of fires burning in the region. Over
the sampling period of this study, over 28 000 hotspots (with a detection
confidence of at least 50 %) were detected, with more than half of these
occurring within 400 km of ATARS (see Fig. 1). For this study, the total
number of observed fires within 10 and 20 km of ATARS was also
calculated (Fig. 2b) for a qualitative assessment of how the smoke from
these fires can affect cloud condensation nuclei concentrations.
Instrumentation
Aerosol size, concentration, composition, hygroscopicity and CCN
concentration measurements were taken to characterise BBA water uptake and
its potential impact on cloud formation. Ambient aerosol was sampled through
an automated regenerating aerosol diffusion dryer to condition the intake to
below 40 % relative humidity. PM1 filters were collected on a TAPI
602 Beta Plus particle measurement system (BAM) for an analysis of EC
and OC. A scanning mobility particle analyzer (SMPS) made up of
a TSI 3071 electrostatic classifier and TSI 3772 condensation particle
counter (CPC) was used to determine the particle size distributions and
number concentrations between 14 and 650 nm with a 5 min averaging
time. A cloud condensation nuclei counter (CCNC) was used to measure total
cloud droplet concentrations at a supersaturation of 0.5 % every 10 s.
A hygroscopicity tandem differential mobility analyser (H-TDMA)
alternated measurement of the hygroscopic growth factor (HGF; D/Dd) of
ambient 50 and 150 nm size-selected particles exposed to a relative humidity
of 90 % (Johnson et al., 2004).
An Aerodyne compact time-of-flight aerosol mass spectrometer (cToF-AMS) was
used to determine the size-resolved chemical composition of non-refractory
sub-micron aerosol. A full discussion on the cToF-AMS analysis of the
composition of bulk PM1 aerosol can be found in Milic et al. (2016).
Briefly, in order to account for fragmentation table issues during periods of
high signals in which some sulfate species were misattributed to organics,
the high-resolution AMS analysis toolkit PIKA was used to separate organic
and sulfate signals. Data for the analysis of the size-resolved chemical
composition within PIKA were not recorded during the sampling period and
therefore the standard AMS analysis toolkit, Squirrel, was used with unit
mass resolution. In order to account for fragmentation table issues related
to the incorrect assignment of organic and sulfate species, the size-resolved
mass concentrations for each species were scaled by the ratio of the mass
concentrations reported by the PIKA analysis to the integrated mass
concentrations reported by Squirrel. The size-resolved composition revealed
that inorganic ammonium and sulfate species made up a greater contribution
of larger particles than in smaller particles (Fig. S1). The composition of
particles between 100 and 200 nm (aerodynamic diameter) was therefore used
in this study as this size range is more representative of aerosols at the
CCN activation diameter. This is further discussed in Sect. 3.3.
Analysis
Total particle number concentrations (PNc) were calculated by integrating
the size distributions measured by the SMPS. The activation ratio of BBA as
CCN at 0.5 % supersaturation was calculated by dividing the CCNc by the
PNc. Apparent activation diameters were calculated by a step-wise
integration of the particle size distribution from the maximum size bin
towards the lower size bins until the total number of particles exceeded the
total number of CCN, as per
CCNc=∫ActivationdiameterUpperdiameterdN/dlogDp⋅dDp,
where CCNc is the total cloud condensation nuclei concentration, N is the
particle number concentration for each size bin, Dp is the particle
diameter, the upper diameter is the largest size measured by the SMPS
(650 nm) and the activation diameter is the size at which the particles
activate to cloud droplets. To calculate the precise activation diameter, a
linear fit (R2>0.98) between the cumulative particle number
concentrations and the diameter was applied across 11 size bins centred on
the bin in which the activation diameter falls. The uncertainty in the
activation diameters was calculated assuming a maximum uncertainty in the CCN
concentrations of ±10 % and was typically of the order of 7 nm. The
extremely vast majority of particles were observed around 100 nm and, on
average, only 0.07 % of the number of particles measured by the SMPS were
between 600 and 650 nm. The influence of particles larger than 650 nm that
were not measured by the SMPS was therefore negligible on the calculation of
CCNc.
The apparent activation diameters were then used to calculate the average
effective hygroscopicity parameter, κ, for each SMPS scan following
κ-Köhler theory (Petters and Kreidenweis, 2007; Petters et al.,
2009). According to this theory, the supersaturation required to achieve a
particular droplet diameter for any given particle can be determined using:
S(D)=D3-Dd3D3-Dd31-κexp4σs/aMwRTρwD,
where S is the supersaturation, D is the droplet diameter, Dd
is the dry particle diameter, κ is the hygroscopicity parameter,
σs/a is the surface tension of the interface between the
solution and air (typically 0.072 J m-2 as pure water is assumed),
Mw is the molecular weight of water, R is the universal gas
constant, T is the temperature (taken as 308 ± 3 K in this study)
and ρw is the density of water. For a range of Dd
values, κ and D values were iteratively varied until the maximum of
the κ-Köhler curve was equal to 0.5 %, the supersaturation
used in the CCNC. A relationship was then found between κ and
Dd for the range of 45 up to 160 nm (Fig. S2). This relationship
was then applied to the calculated activation diameters over the sampling
period to calculate the BBA κ values. The uncertainty in the
activation diameter of 7 nm led to uncertainties in κ of
∼ 0.05 for activation diameters between 60 and 80 nm, ∼ 0.01 for
activation diameters between 80 and 100 nm, and less than 0.01 for activation
diameters above 100 nm.
The value of κ derived from the CCNC and SMPS is the average for all
particle sizes. If there is not a uniform composition, this value cannot
necessarily be applied to all sizes of BBA. All H-TDMA data were inverted
using the TDMAinv algorithm (Gysel et al., 2009) and HGF distributions were
kelvin-corrected for comparison between 50 and 150 nm particles at 90 %
relative humidity. Equation (2) was then also applied to the kelvin-corrected
HGF distributions, thereby providing distributions of κ. This also
provides an insight into the mixing state of the BBA, which cannot be
determined from the CCNC and SMPS measurements in this study.
When the surface tension of pure water is assumed, κ is regarded as
the “effective hygroscopicity parameter”, which accounts for changes in
water activity due to the solute as well as any surface tension effects
(Petters and Kreidenweis, 2007; Rose et al., 2010; Pöschl et al., 2009).
The effective hygroscopicity parameter is therefore an indication of all
compositional effects of an aerosol particle on water uptake. To distinguish
the potential effects of surface tension, values of 0.052 and
0.0683 J m-2 were also applied. Mircea et al. (2005) suggest that the
surface tension at the liquid–air interface of a particle depends on the
concentration of carbon. The value 0.052 J m-2 represents a lower
limit, while a surface tension of 0.0683 J m-2 has been observed for
prescribed biomass burning particulate matter in wooded areas in the USA
(Asa-Awuku et al., 2008). The impact of surface tension is discussed further
in Sect. 3.5.
The overall hygroscopicity of any given particle can be determined by the
volume fraction and hygroscopicity of each constituent under the Zdanovskii,
Stokes and Robinson (ZSR) assumption (Chen et al., 1973; Stokes and Robinson,
1966):
κ=∑iεiκi,
where κ is the overall hygroscopicity and εi and
κi are the volume fractions and hygroscopicities of each
constituent, respectively. A modelled κ was constructed to determine
the influence of diurnal changes in organic and inorganic volume fractions,
where κtotal=εorgκorg+εECκEC+εinorganicκinorganic, following the ZSR
assumption. The 12 h PM1 BAM filters sampled from 07:00 until 19:00 and
from 19:00 until 07:00 each day showed no difference in the ratio of EC to
(OC + EC), and therefore a constant mass fraction (EC / (EC + OC)
of 10 % was applied. εorg, εEC
and εinorganic were calculated using the size-resolved
mass concentrations reported by the cToF-AMS and assumed densities of
1.4 g cm-3 (Levin et al., 2014), 1.8 g cm-3 (Bond and
Bergstrom, 2006) and 1.8 g cm-3 (Levin et al., 2014), respectively.
κinorganic and κEC were taken as 0.60
(Bougiatioti et al., 2016) and 0 (Petters and Kreidenweis, 2007),
respectively. The contribution of inorganics in this study was taken from
the reported masses of sulfate, nitrate and ammonium species. During the
period considered for this model, other inorganic species such as potassium,
a marker for biomass burning, only made up a small and constant contribution
to the total mass and were therefore not considered. A night (18:00 until
07:00) value and a day value of κorg were varied in the applied model
in order to investigate any potential changes in organic hygroscopicity due
to photochemistry.
CCN concentrations were calculated in order to test prediction of CCN
concentrations using aerosol composition and size distribution in this
region. Activation diameters were derived from various hygroscopicity
parameters and, again using Eq. (1), the size distribution was integrated
step-wise from the upper size-limited measured in the SMPS until the
activation diameter was reached. The same process used to calculate the
precise activation diameters earlier was used to calculate the precise CCN
concentrations. This process was carried out for the modelled hygroscopicity
from the size-resolved cToF-AMS data, the measured hygroscopicity
distribution from the H-TDMA and various constant hygroscopicity
values. The constant values selected were 0.05, 0.1, 0.2 and 0.3, as well as day and
night values of 0.071 and 0.035, respectively; 0.05 represents the campaign
average effective hygroscopicity. The day and night values represent the
campaign average values obtained from the SMPS-CCNC measurements. The values
of 0.1 and 0.2 represent commonly observed hygroscopicities for BBA in other
regions and in laboratory measurements (Engelhart et al., 2012). The global
mean values of κ have been estimated to be 0.27 ± 0.21 for
continental aerosols (Pringle et al., 2010). It has been suggested that it is
suitable to assume this continental average (κ∼ 0.3) to make
first-order predictions of CCN activity (Rose et al., 2011). Modelling CCN
concentrations using these methods and assumed hygroscopicity values will
verify whether such values are suitable in predicting CCN activity in regions
like northern Australia.
The time series of total cloud condensation nuclei concentrations
(CCNc) at 0.5 % supersaturation and the total number of
satellite-observed fires within 10 km (yellow) and 20 km (red) of the
sampling location.
The diurnal trends of (a) the total cloud condensation
nuclei concentration (CCNc) and particle number concentration (PNc),
(b) the activation ratio at 0.5 % supersaturation,
(c) the median and mode of the particle size distribution, and
(d) the apparent activation diameter. All reported values are the
median of the hourly averaged data for the sampling period and the error bars
represent the standard deviation.
In order to investigate the CCN activity of BBA, 4 days of unpolluted and
coastal conditions (19–22 June 2014) were removed from the majority of the
analysis. Furthermore, the SMPS was only operational from 4 June 2014.
Analysis of CCNc, PNc, activation ratios, median particle diameters, apparent
activation diameters and the average effective hygroscopicity parameters is therefore only presented for data collected after this date.
Results and discussion
BBA contribution to CCN
Figure 2 shows the CCN concentrations (supersaturation 0.5 %) measured at the ATARS
over the campaign sampling period in June 2014 as well as the frequency of
fires that were observed via satellite hotspots each day within 10 and 20 km
of the station. Air mass back trajectories were typically from the southeast,
as were the locations of the fires (Mallet et al., 2016). These back
trajectories revealed that the air masses did not pass over Darwin or any
close industrial sites, ruling out the likelihood of an urban influence on
CCN concentrations. The period between 19 and 23 June was characterized by
relatively low CCN concentrations due to air originating from the coastal
waters of eastern Australia, which passed over minimal continental area
before arriving at the ATARS. As already mentioned, these dates were
subsequently excluded from the data analysis as the focus of this study was
on the impact of BBA on CCN. The highest PNc and CCN concentrations were
associated with large biomass burning events. PNc concentrations of up to
400 000 cm-3 and CCN concentrations of up to 19 000 CCN cm-3
were observed during these periods.
The diurnal trends of (a) the CCNC-derived effective
hygroscopicity parameter, (b) the H-TDMA-derived kelvin-corrected
hygroscopicity distributions of 50 nm particles, (c) the
AMS-derived hygroscopicity parameter for aerodynamic diameters between 100
and 200 nm, assuming κorg of 0.02 and 0.08 during the night
and day, respectively, and (d) the H-TDMA-derived kelvin-corrected
hygroscopicity distributions of 150 nm particles. The black dots
in (b) and (d) represent the hourly median hygroscopicity
values and the error bars represent the standard deviation.
Although the PNc and CCN concentrations were highest during BBA events, these
periods were characterized by the lowest hygroscopicity and activation ratios
(ratio of CCNc to PNc as low as 4 %). This is further discussed in
Sect. 3.3. Activation ratios typically varied between 30 and 80 %,
corresponding to CCN concentrations between 1500 and 6000 cm-3. This
contrasts with observed activation ratios of over 80 % in BBA from dry season
savannah fires in tropical southern Africa during the SAFARI 2000 campaign
(Ross et al., 2003), despite lower supersaturations of ∼ 0.3 %. The
size distributions of BBA observed during SAFIRED had a count median diameter
of 107 ± 25 nm, while the median diameters were typically above
150 nm in the SAFARI 2000 campaign, which could explain the lower activation
ratios observed here. The size distributions observed in this study were
typically smaller than those observed in aged and regional BBA on other
continents (Reid et al., 2005). When particles are smaller, the critical
diameter for cloud droplet formation becomes more important. It is therefore
crucial to investigate the impact of composition on the activation diameter,
and thus CCN concentrations.
Diurnal trends in BBA
Diurnal patterns in the BBA PNc, CCNc, size and activation ratio and
activation diameter are shown in Fig. 3. For most of the campaign, particle
size distributions were unimodal and therefore the median and mode of these
distributions are used here to represent the particle size. The highest
concentrations of CCN were observed during the night, when they were also the
most variable (Fig. 3a). This is likely a result of prescribed burns
occurring later in the day or evening as well as a lower inversion layer
during the night. Interestingly, the activation ratio also follows a distinct
diurnal trend with ∼ 40 ± 20 % of BBA acting as CCN at
0.5 % supersaturation during the night and ∼ 60 ± 20 %
during the day (Fig. 3b). Smaller particles were typically seen during the
day than during the night (Fig. 3c), indicating that it was the change in the
particle activation diameter (Fig. 3d) that was responsible for this increase
in daytime activation ratios. The decrease in the particle size could
possibly be explained by changes in combustion (Carrico et al., 2016) of
vegetation across the day, with more flaming, rather than smoldering,
conditions expected to be favoured during daylight hours. Without more
information on the exact location, fuel type and combustion conditions,
however, it is difficult to make conclusions about this.
The hygroscopicity of BBA derived from the size-resolved AMS, CCNC/SMPS and
the H-TDMA followed a distinct diurnal trend (Fig. 4). The CCNC-derived
hygroscopicity (Fig. 4a) during the night-time (defined as 18:00 until 07:00
local time) was generally very stable and constant over the sampling period
at 0.03 ± 0.03. Daytime hygroscopicity (07:00 until 18:00) was
typically higher with much more variability at 0.07 ± 0.05.
H-TDMA-derived hygroscopicities for 150 nm diameter particles (Fig. 4d)
agreed very well with these values, although a much higher variability was
observed around noon. The hygroscopicity distributions of 50 nm diameter
particles followed a similar trend but were, interestingly, slightly higher
than the 150 nm distributions. The variability in hygroscopicity for 50 nm
diameter particles is much greater than for 150 nm particles, due to lower
concentrations at 50 nm. Hygroscopicity distributions for both 50 and
150 nm aerosols during the night indicate a strong internal mixture of very
weak hygroscopic BBA, and during the day an increase and broadening of the
hygroscopicity mode, suggesting an external mixture of slightly more
hygroscopic particles. The size-resolved AMS hygroscopicity values were
calculated assuming κorg of 0.02 and 0.08 during the night and
day, respectively. The organic volume fraction was invariable (see
Sect. 3.3); therefore, the increase and decrease at sunrise and sunset,
respectively, are driven by the choice of night and day organic hygroscopicity
values. These values were selected as they gave the best agreement between
the modelled and measured CCN concentrations, which is discussed further in
Sect. 3.4.
Literature on the diurnal variability in BBA hygroscopicity is rare. Diurnal
and afternoon averages of κ for BBA in the Amazonian dry season have
been reported as 0.048 and 0.072, respectively (Sánchez Gácita et
al., 2017), consistent with the results presented here. A short study
(Fedele, 2015) carried out in 2010 at the ATARS also reported κ
values in the early dry season over a period of 2 weeks. They showed
κ values mostly between 0.05 and 0.1 for supersaturations of 0.38,
0.68 and 0.96 %, with the higher values generally occurring during the
day. They directly measured the critical diameter and used an approximation
presented in Petters and Kreidenweis (2007) to calculate κ. This
approximation is more appropriate for κ values over 0.2, which means
that the reported values between 0.05 and 0.1 were slightly overestimated and
would likely be more in line with the BBA hygroscopicity observed in this
study. Although a detailed chemical analysis was not done during that study,
these similar values of κ suggest that these observations could be
representative of early dry season fires in this region.
BBA composition
The activation ratio as a function of the effective hygroscopicity parameter,
κ, as calculated from Eqs. (1) and (2), with colours indicating the
median particle mobility diameter, is displayed in Fig. 5. This figure
clearly demonstrates that both the size and composition of the BBA can have a
significant effect on CCN activation. For example, with a constant particle
size increases the CCN activation ratio from below 20 to above 80 %. For
a constant κ of 0.05 and an increase in the particle median diameter
from 60 up to 140 nm, the CCN activation ratio increases by approximately
50 %. The effect of composition appears to have less of an influence at
higher hygroscopicities, with the size being the determining factor in CCN
activation above a κ of 0.1. For very weakly hygroscopic (κ<0.05) BBA, the sensitivity of particle size was less prominent, with an
activation ratio increase of ∼ 0.3 % nm-1, compared to a
∼ 0.7 % nm-1 increase when κ>0.05. These findings
support the idea that cloud droplet number concentrations are sensitive to
composition at low hygroscopicities (Reutter et al., 2009). Neglecting the
effect of BBA composition in this case would lead to difficulties in
appropriately quantifying CCN activation.
The activation ratio (CCNc / PNc) as a function of the effective
hygroscopicity parameter, κ. The colours represent the median
particle mobility diameter.
In order to understand the underlying causes of the variations in the
hygroscopicity, the size-resolved chemical composition was investigated. For
bulk PM1 composition, there was a distinct increase in the inorganic
mass fractions during the day due to an enrichment of ammonium and sulfate
species. The size-resolved composition, however, revealed that these
inorganic species were more present on larger particles and had a
daerodynamic mode at approximately 350 nm, while organics had a
mode at approximately 250 nm. This observation is consistent with other
studies that show that smaller particles are more enriched with organics
(Levin et al., 2014; Rose et al., 2011; Gunthe et al., 2009). As the
influence of composition on CCN activation is irrelevant at larger sizes, it
is important to investigate the composition at smaller sizes where the
aerosol number is highest and the composition can affect the activation
diameter. The size-resolved composition revealed that, within the aerodynamic
diameter size range of 100 to 200 nm, organics were completely dominant and
the organic volume fraction, εorg, was invariable at
approximately 90 %. At least some of the increase in the observed
hygroscopicity, κ, and the inferred κorg is likely a result of the
photochemical oxidation of the organics. The aging of biomass burning aerosol
is discussed in further detail in Milic et al. (2016), where the fraction of
m/z 44 to total organics measured by the AMS, a proxy for the degree of
oxidation, was shown to increase steadily throughout the day. As shown in
Fig. 4, the derived hygroscopicity values from the CCNc, SMPS and H-TDMA show
a decrease in the hygroscopicity soon after the peak at midday. If the
photochemical oxidation of organics were the sole contributor to the daytime
increase in hygroscopicity, then it should be expected that the hygroscopicity
would also increase steadily throughout the day (in the absence of a change
in the mass fraction of inorganics). While there was no change in wind
direction until later in the afternoon, the peak in hygroscopicity did
correspond with the peak in wind speed (see Fig. S4 in the Supplement),
although it is not apparent how or if a decrease in the wind speed could lead
to a decrease in the hygroscopicity. A separate explanation could be related
to the size-dependent composition of BBA. The size-resolved composition from
the AMS across the range of 100–200 nm was selected due to the inefficient
transmission of particles below 100 nm. As shown in Fig. 3d, the apparent
activation diameter during the day decreased to approximately 80 nm. It
could be that the composition between 100 and 200 nm is therefore not
perfectly representative of the BBA at the activation diameter. Furthermore,
the influence of other inorganics not considered in the model of
hygroscopicity or the role of surface chemistry could be underestimated,
leading to poor characterisation of hygroscopicity by bulk composition.
The CCNc, activation ratio, median particle diameter, apparent
activation diameter and average effective hygroscopicity parameter during two
periods with close-proximity (< 1 km) fires. The green shaded area
indicates the period where the wind direction was periodically changing
between southeasterly and northeasterly. The red shaded area indicates the
period where emissions were from a grass fire burning less than 1 km from
the sampling site.
While it is likely that most of the BBA observed during the SAFIRED campaign
had undergone some form of aging (physical or chemical), two events provided
insight into the characteristics of extremely fresh BBA and are described in
Mallet et al. (2016). During mid-afternoon on 25 June, grass and shrub fires
were blazing ∼ 1 km southeast of the ATARS site. Wind directions
during this period were very unstable and frequently altering between
southeasterly and northeasterly. This resulted in the sampled air mass
frequently changing from the “fresh” plume and more background-like
conditions over the course of approximately 4 h. During this time, CCN
concentrations varied frequently between ∼ 2000 and
∼ 19 000 cm-3. The activation ratio, median particle size,
activation diameter and hygroscopicity varied between 20 and 75 %; 80 and
110 nm; 130 and 80 nm; and 0.02 and 0.1, respectively (see Fig. 6). These fires
continued to blaze into the evening, slowly advancing to within 1 km south
of the ATARS site. Due to northeasterly winds, the air mass from this fire
was not observed until approximately 22:00 that night, when winds became
southerly. For the next 4 h, CCN concentrations peaked at
∼ 19 000 cm-3, despite the activation ratio dropping to 4 %.
The average effective hygroscopicity during this event dropped to 0.003 and
slowly increased over the period of the fire to ∼ 0.02. This led to a
decrease in the apparent activation diameter from 250 to 150 nm,
subsequently increasing the CCN activation ratio to 25 %. Whether this is
a result of a change in the burning conditions, fuel load or a combination of
both is unclear. These events demonstrate the importance of BB as a source of
CCN, despite the relatively hydrophobic nature of BBA. Furthermore, in the
absence of photochemical aging, the slight variation in BBA hygroscopicity
during the night fire demonstrates the variability in CCN activation, even
over the course of an individual fire.
The hygroscopicity of fresh and aged BBA has been studied extensively in
laboratory smog chambers. Some studies have shown that the photochemical
oxidation of organics in BBA leads to an increase in hygroscopicity (Carrico
et al., 2010; Petters et al., 2009; Engelhart et al., 2012), while others
suggest that the hygroscopicity converges from highly (κ=0.6) or weakly (κ=0.06) hygroscopic values to a value of
approximately 0.2 ± 0.1 (Engelhart et al., 2012). The observation of
the close-proximity fire event on the evening of 25 June (Fig. 6), as well as
the diurnal trends in the calculated hygroscopicity parameter (Fig. 4),
indicates that the composition of BBA during the night is characteristic of
very weakly hygroscopic fresh BBA. The high frequency of fires during the
early dry season in northern Australia likely results in the “regional haze”
predominantly being composed of relatively fresh BBA with a very low
hygroscopicity. The aging processes were observed to increase the
hygroscopicity to ∼ 0.08 ± 0.05, which is the lower estimate of
value suggested by Engelhart et al. (2012) and smaller than other studies
investigating BBA (Bougiatioti et al., 2016). Whether the hygroscopicity
would converge to higher values in the absence of frequent fires or as the
smoke travels away from the continent is something that needs to be explored
in future measurements.
The normalised probability density functions for the frequency of
the ratio of the modelled and measured CCN concentrations based on
(a) the 150 and 50 nm H-TDMA-derived hygroscopicities and
size-resolved cToF-AMS composition and (b) the campaign average
hygroscopicity, κ=0.05; typical BB hygroscopicities of 0.1 and 0.2;
the continental average hygroscopicity of 0.3; and the day and night average
effective hygroscopicity values of 0.071 and 0.035, respectively.
Validation of modelled CCN
Detailed temporal–spatial measurements of CCN concentrations are difficult
and therefore assumptions must be made about the size-resolved composition
and water uptake for similar regions and sources. Many studies have attempted
closure between composition, size and CCN concentrations in order to assess
the validity of these assumptions (Rose et al., 2010). These studies
typically agree that for most environments, where hygroscopicities are
moderate, the size distribution and number concentration of particles are the
determining factors of CCN concentrations (Dusek et al., 2006).
Sánchez Gácita et al. (2017), however, showed that for Amazonian BBA (with
measured κ=0.04), applying an assumed κ of 0.20 resulted in
a 26.6 to 54.3 % overestimation of CCN concentrations. They suggest that
κ values recommended for continental and BBA are too high to describe
CCN behaviour of Amazonian BBA. This is also the case for the SAFIRED
campaign.
Figure 7 shows the normalised frequency distributions of the ratio of
modelled and measured CCN concentrations for five different compositional
scenarios, taking into account the time-dependent size distributions. The
smoothed curves were obtained frequency distributions of these ratios, using
Igor Pro's Multi-peak Fitting package. For a constant hygroscopicity of 0.20,
daytime concentrations were overestimated by 15 to 40 %, while night
concentrations were overestimated by well over 100 %. A similar case is
observed for hygroscopicities of 0.10 and 0.30. For an assumed constant
κ of 0.05, which represents the campaign average, the modelled CCN
concentrations slightly underestimate the measured CCN concentrations during
the day by less than 10 % but overestimate the night CCN by 65 %.
Using the day and night campaign averages of 0.071 and 0.035, respectively,
improved the night-time concentration to an overestimation of approximately
50 %. Using the time-dependent size-resolved AMS composition and
assigning and κorg as 0.08 and 0.02 for day and night,
respectively, also provides a good agreement between the estimated and
measured daytime CCN concentrations, but it again overestimates the night
concentrations by 70 %. H-TDMA hygroscopicity distributions showed that
the night was predominantly characterised by an internal mixture, suggesting
that the disagreement between modelled and measured CCN is due to is due to
variability in fuel or burning conditions of the fires and/or night-time
aging. The modelled CCN concentrations from both the 50 and 150 nm H-TDMA
were within 10 % during the day, but there were also overestimations of
between 45 and 70 %, respectively. The time resolution of the H-TDMA
limited the number of CCN model calculations that could be done, which
introduced more potential bias for individual periods where the agreement
between the measured and modelled CCN was worse (or better). The difficulty
in sufficiently modelling night-time CCN concentrations highlights the need
for further composition measurements of fresh BBA in this region.
Effect of surface tension
The κ values reported in this study represent the effective
hygroscopicity parameter, which accounts for all compositional effects on
aerosol water uptake (i.e. solubility of components and the reduction in
surface tension to their presence). This study has shown that the effective
hygroscopicity parameter increases during daylight hours, leading to speculation that
this is caused by the photochemical oxidation of organics. Although surface
tension measurements were not performed, using a value observed in a previous
BB study of 0.0638 J m-2 (Asa-Awuku et al., 2008) in the κ-Köhler equation shows only a slight decrease in hygroscopicity
compared to using the surface tension of pure water (Figs. S2 and S3). This
suggests that it is the solubility, rather than the reduction of surface
tension, of the organics and inorganics present in the BBA that is
responsible for the water uptake. On the other hand, an assumed lower
estimate surface tension of 0.052 J m-2 (Mircea et al., 2005) during
the day could explain the increase in CCN activation. Although models
generally use the effective hygroscopicity parameter (Pringle et al., 2010)
due to the ease of using a single parameter, a better understanding of the
precise mechanisms that facilitate the uptake of water onto potential cloud
droplets is needed (Noziere, 2016).
Conclusions
Measurements at ATARS showed a strong link between the frequency of early dry
season fires and the concentrations of CCN, indicating that these fires are
an important source of CCN in northern Australia. The aerosol size
distribution was typically unimodal with a median diameter of 107 nm and the
BBA was weakly hygroscopic and predominately internally mixed. These
conditions meant that both the composition and size were important in
determining the CCN activation of the BBA. A distinct diurnal trend in the
ratio of activated cloud condensation nuclei at 0.5 % supersaturation and
particle number was observed, with ∼ 40 ± 20 % of BBA acting
as CCN during the night and ∼ 60 ± 20 % during the day. This
increase in CCN activity corresponded with an increase in the hygroscopicity
from 0.04 ± 0.03 to 0.07 ± 0.05. This was likely due to the
daytime photochemical oxidation of organic compounds within BBA, although
other factors such as changes in sub-100 nm inorganic contributions or
surface chemistry were likely also contributing factors. While not
investigated in this study, this smoke has the potential to penetrate into
the upper levels of the troposphere, particularly as the dry season
progresses, and it also flows over the Timor Sea, where change in cloud albedo
and lifetime is likely to be sensitive to CCN concentration changes. In the
case of northern Australian dry season fires, assuming typical continental
hygroscopicities of 0.10, 0.20 and 0.30 led to CCN overestimates of 10 to
30 % during the day and 100 to over 150 % during the night. It is
therefore important that the CCN activation be better modelled.
BBA related CCN concentrations are likely to be further enhanced throughout
the dry season as temperatures increase and there are more frequent fires.
Long-term monitoring or future measurements later in the dry season would
allow a more detailed analysis into the seasonal relationship between fire
frequency, intensity and CCN. Other aerosol–cloud interactions are likely to
change as the season progresses. Higher solar radiation and relative
humidity during the late dry season lead to the formation of pyro-cumulous
clouds and higher rainfall in comparison to the early dry season (Bowman
et al., 2007). Long-term measurements could also be further integrated in
with satellite data, such as the burned-area product measured by the MODIS
sensors. While the number of fires detected in this study is useful for a
qualitative assessment of the impact of fires on CCN concentrations, the
area burned is likely to be a more quantitative proxy for BBA emissions.
Concurrent aircraft measurements would be required to investigate the
penetration and evolution of smoke into upper levels of the troposphere.
Characterising the presence of smoke within, below and above clouds is
required to fully understand the vertical radiative effect of these fires.
The southeasterly trade winds carry this smoke over waters in the Indian
and western Pacific oceans known as the tropical warm pool. Measurements in
Indonesia or on a ship in the Timor Sea would therefore also be useful in
determining the long-range transport and evolution of the smoke.
Furthermore, a mobile sampling chamber positioned downwind of prescribed
burns that occur in this region, both during the day and night, would be
beneficial in understanding the variability in the composition of freshly
emitted BBA.