Introduction
The significant impact of black carbon (BC) aerosol on radiative forcing is
well established (Bond et al., 2013), but the magnitude and wavelength
dependence of absorption by organic carbon (often called brown carbon, BrC)
remains poorly constrained (Barnard et al., 2008; Lack et al., 2012a;
Kirchstetter and Thatcher, 2012; Bahadur et al., 2012; McMeeking et al.,
2014; Laskin et al., 2015; Olson et al., 2015). If BC is coated by
non-absorbing organics (primary or secondary organic aerosol) or inorganic
non-absorbing materials (ammonium sulfate, ammonium nitrate, etc.) these
coatings can enhance the magnitude of absorption by the BC core. This effect
is often called, somewhat inaccurately, lensing, and for simplicity this term
will be utilized in this paper (Fuller et al., 1999). The lensing effect is
slightly decreased if the coatings themselves are absorbing (Lack and Cappa,
2010). While Mie calculations and lab studies support the notion that
significant (50–100 %) increases in BC absorption can occur via lensing at
atmospherically relevant aerosol sizes, observations of this effect in
ambient air have yielded a wide range of results (Bond et al., 2006; Cappa
et al., 2012; Nakayama et al., 2014). Studies have also found a wide range
of possible imaginary refractive indices for BrC (Kirchstetter et al., 2004;
Lack et al., 2012a; Alexander et al., 2008; Saleh et al., 2014), with Saleh
et al. (2014) postulating that the imaginary refractive index can be
predicted if the ratio of black carbon to organic aerosol mass (BC : OA) is known.
Potential sources of BrC include emissions from biomass burning
(Kirchstetter et al., 2004; Moosmüller et al., 2009; Chen and Bond, 2010;
Lack et al., 2012a; Saleh et al., 2014; Washenfelder et al., 2015);
incomplete combustion of fossil fuels, especially coal (Bond, 2001; Yang et
al., 2009; Olson et al., 2015); and secondary organic aerosols (Saleh et
al., 2013; Zhang et al., 2013; Liu et al., 2014; Lin et al., 2015). There
exists significant uncertainty concerning the relative contribution of each
of these source types to total BrC concentrations, but several studies have
identified biomass burning as a potentially significant source (Washenfelder
et al., 2015; McMeeking et al., 2014; Lack et al., 2012a). Open biomass
burning (BB) is one of the largest global sources of BC and
organic carbon (OC), and biomass burning emissions have a significant direct
effect on the Earth's radiative balance (Bond et al., 2013). When biomass
burning emissions interact with clouds, there are significant semi-direct
and indirect effects, and the magnitude of the semi-direct effects depends on
the optical properties of the emitted aerosols (Sakaeda et al., 2011; Lin et
al., 2014; Jacobson, 2014).
Absorption enhancement can be defined as the ratio of the absorption of
ambient aerosol (with coatings present) to the absorption of BC alone.
Experimental attempts to quantify absorption enhancement often employ
thermal denuders to remove organic and inorganic coatings. Lack et al. (2012a)
measured ambient aerosols with a large influence of biomass burning
and found absorption enhancements up to 2.5 at 405 nm and 1.4 at 532 nm
using a thermal denuder. Laboratory measurements of absorption enhancement
in biomass burning aerosols show significant variation with values greater
than 2 observed at 405 nm in some samples and values ranging from
1.2 to 1.5 at 532 and 781 nm (McMeeking et al., 2014). There are a number of studies
that estimate the absorption due to BrC in the blue and ultraviolet
wavelengths (Yuan et al., 2016; Washenfelder et al., 2015; Guo et al., 2014;
Nakayama et al., 2014; Lack et al., 2012a; Cappa et al., 2012; Bahadur et
al., 2012; Flowers et al., 2010; Wang et al., 2016; Stockwell et al., 2016).
Results from these studies show large variations in the estimated percentage
of absorption attributable to BrC. Laboratory and field studies have
measured wide ranges of BrC absorptivity values (Kirchstetter et al., 2004;
Lack et al., 2012a; Alexander et al., 2008; Saleh et al., 2014). Recent
model studies have shown that inclusion of BrC could significantly alter the
direct radiative forcing due to carbonaceous aerosols (Saleh et al., 2015;
Feng et al., 2013). Additionally, BrC is thought to lose its absorptivity in
a relatively short (hours to days) amount of time aging in the atmosphere
(Wang et al., 2016; Forrister et al., 2015; Lee et al., 2014). This aging
provides another reason it is critical to differentiate between absorption
from BrC and BC.
To understand the relative importance of absorption from BrC vs. lensing
and BC, it is critical to understand the potential variability in
attribution of absorption caused by different methodologies. Various studies
show large variations in the percentage of absorption due to BrC (Yuan et
al., 2016; Washenfelder et al., 2015; Guo et al., 2014; Nakayama et al.,
2014; Lack et al., 2012a; Cappa et al., 2012; Bahadur et al., 2012; Flowers
et al., 2010) at blue wavelengths. While some of the variation is certainly
due to variations in the ambient aerosol, some could potentially be the
result of various approaches used to estimate the contribution of BrC to
total light absorption in previous studies. Lack et al. (2012a) and Saleh et
al. (2014) used core–shell Mie theory with inputs of BC core and shell size
and the imaginary refractive index of BrC. Particle morphology is also taken
in account, using the Rayleigh–Debye–Gans approximation, in some studies
(Liu et al., 2015). Others have used the difference in absorption
enhancement from thermally denuding the aerosol at different wavelengths to
determine the contribution from evaporated BrC (Guo et al., 2014;
Nakayama et al., 2014; Cappa et al., 2012), with the assumption that the
lensing effect depends weakly on wavelength (Lack and Cappa, 2010). Another
method that is simple and widely applied assumes an absorption Ångström
exponent (AAE) value for BC (typically 1) and defines absorption above the
predicted BC absorption at low wavelengths to be BrC, while sometimes also
accounting for lensing (Fialho et al., 2005; Favez et al., 2009; Bahadur et
al., 2012; Cazorla et al., 2013; Yuan et al., 2016). In this study, we
estimate the percentage of absorption due to BC, lensing, and BrC following
methodologies based on assuming an AAE and based on thermally denuding the
aerosol, thereby providing a range that covers variations caused by
experimental approach, of potential contributions to absorption in biomass
burning emissions from BrC, lensing and BC.
Recent studies have shown that certain parameters can be used to
parameterize biomass burning aerosol optical properties. These parameters
include modified combustion efficiency (MCE), black carbon to organic
aerosol ratio (BC / OA), and elemental to total carbon ratio
(ECEC+OC) (Pokhrel et al., 2016; Lu et al., 2015; Saleh et al.,
2014; McMeeking et al., 2014; Liu et al., 2014). The current study
demonstrates that the magnitude of total BrC absorption can be
parameterized with the EC / OC ratio, a distinct result from Saleh et al. (2014) who
demonstrated that the intensive quantity of absorptivity, or imaginary
refractive index, can be parameterized with the BC / OA ratio.
Materials and methods
Measurements were made during the multi-investigator FLAME-4 experiment.
Details of the fuels burned and overall experiment can be found in Stockwell
et al. (2014), and details of the experimental setup for optical measurements
in Pokhrel et al. (2016). The schematic of the instrumental setup during the
FLAME-4 experiment is shown in Fig. 1. Briefly, this study reports results
from 12 different fuels with significant global emissions over 22 individual
burns, as detailed in Fig. 4. All results presented are from burns where a
fire was ignited in a large combustion room at the Missoula Fire Sciences
Laboratory and allowed to burn to completion. Measurements were made after
smoldering had ceased and emissions from all stages of the fire had become
well mixed in the combustion room.
Inlet system
Aerosol was pulled from an inlet placed roughly 3 m above the floor of
a 12.5 m × 12.5 m × 22 m combustion room once the smoke
was well mixed (typically 15–20 min after the start of a burn). Fuels
burned include ponderosa pine, black spruce, rice straw, organic hay,
organic and conventional wheat, sugarcane, giant cut saw grass, wire grass,
African grass, chamise, manzanita, and North Carolina peat. Smoke was
transferred to the instruments at 10 L min-1 through a 0.5 inch
outside diameter (OD) copper line. Aerosol was passed through a cyclone impactor that removed
particles with aerodynamic diameters larger than 2.5 µm and diluted
with particle-free air in order to maintain extinction levels near
500 Mm-1 to avoid saturation of sampling instrumentation. It is important
to note that aerosol sampled for EC / OC measurements (Sect. 2.7) was not
diluted. Next, the air was dried by two 100-tube Nafion driers (Perma Pure,
Toms River, NJ, USA) operated in parallel, which reduced the relative humidity in
the sample cell to less than 15 %. Following the Nafion driers, an
activated carbon monolith (MAST Carbon, Basingstoke, UK) was used to scrub
NOx and O3 from the sample air while transmitting the particles.
The successful removal of NOx was continuously tracked by a cavity ring-down spectrometer (CRDS) gas-phase channel at 405 nm. A filter was
periodically inserted (at an interval of 5 to 10 min) into the sample
stream to remove particles and confirm baseline stability.
The schematic of instrumental setup during the FLAME-4 experiment.
Instrumentation
Aerosol absorption measurements were made with a multi-wavelength
photoacoustic absorption spectrometer (PAS) instrument (Lack et al., 2006;
2012b). The PAS was configured with five cells that measured absorption
coefficients of dry aerosol (RH < 15 %) at 405, 532, and 660 nm
and denuded aerosol at 405 and 660 nm. An integral component of the PAS was
a thermal denuder system. The thermal denuder deployed during FLAME-4 was a
50 cm long, 1.1 cm inside diameter (ID) stainless steel tube heated to 250 ∘C
followed by a 50 cm 0.43 in. ID unheated tube with an activated carbon
honeycomb monolith (MAST Carbon, Basingstoke, UK). The denuder was run at
250 ∘C in an attempt to avoid charring of organic aerosol which would
artificially increase absorbing material in the denuded channel. The flow
rate through the thermal denuder was 5 L min-1. The activated carbon section
absorbed volatile species that evaporate from the aerosol in the heated
region. This denuder configuration was shown to have almost complete removal of a semi-volatile organic compound
(di-ethyl-hexyl-sebacate having a boiling point of 300∘ C) coating
sodium chloride cores (Fierz et al., 2007). It is known that biomass burning
emissions often have low-volatility material and these materials can be
absorbing (May et al., 2013; Saleh et al., 2014). It is possible, in fact
probable given that the denuder was run at 250 ∘C to avoid charring,
that all of the absorbing organic material was not removed by the thermal
denuder in this study, which will be discussed in detail in the results and discussion.
Particles loss through the thermal denuder as a function of particle
size and thermal denuder temperature.
PAS calibration
A CRDS (Langridge et al., 2011) that
operates at identical wavelengths to the PAS (660, 532, 405 nm) was
used to calibrate PAS absorption measurements. Calibrations were conducted
using ozone sampled simultaneously by the PAS and CRDS. PAS absorption at a
given wavelength was set equal to the measured extinction from the CRDS at
that same wavelength, with Rayleigh scattering subtracted. Further details of
the PAS calibration using the CRDS can be found in Lack et al. (2012b).
Further details about the instrumental setup of the PAS and CRDS can be
found in Pokhrel et al. (2016). The uncertainty associated with the PAS
measurements during FLAME-4 was found to be 7 % (1σ) in the
660 and 532 nm channels and 18 % (1σ) in the 405 nm channel (Pokhrel
et al., 2016). The higher uncertainty in the 405 nm channel is due to the
use of ozone during PAS calibration. The absorption cross section of ozone
at 405 nm is 1.51 × 10-23 cm2 molecule-1 (Axson et al.,
2011), which is 2 orders of magnitude less than the absorption cross
section in the 450–858 nm range (10-21 cm2 molecule-1) (Anderson and
Mauersberg, 1992). During ozone calibration, the absorption coefficients
at 405 nm were always less than 20 Mm-1, which resulted in a
calibration curve with significantly more uncertainty than the curves for
the other wavelengths when extrapolated to the high absorption values observed in this study.
Particles loss in thermal denuder
Particle losses occur inside the thermal denuder because of thermophoretic
and diffusional processes (Wehner et al., 2002). Particle loss in the
thermal denuder was measured by passing glassy carbon spheres through the
denuder at a flow rate of 5 L Min-1. Different-sized particles were passed
through the thermal denuder at multiple temperatures. Figure 2 shows the
percentage of particles that were transmitted through the thermal denuder.
The percentage of particles lost in the thermal denuder depends on both
particle size and thermal denuder temperature. During FLAME-4, the vast
majority of particles for all burns were less than 300 nm (Saleh et al.,
2014) and the temperature of the thermal denuder was set to 250 ∘C.
Under these conditions, the transmission efficiency of the thermal
denuder is 71 % on average. Based on this, denuded absorption was divided
by 0.71 to correct for losses. It is important to note that this size range
and temperature result in nearly the maximal particle loss through the
denuder, observed in Fig. 2, and it is therefore possible that particle
losses through the denuder could be overcorrected. If this occurs, it will
lead to an underestimate of absorption enhancement in particles. The
efficiency curves are relatively insensitive to diameter, making it a
reasonable assumption to have a fixed efficiency factor and not have to
consider complications associated with particle size, and thus transmission
efficiency, changing as the particle evaporates on passing through the
denuder. Absorption enhancements based on the thermal denuder measurements
as low as 0.91 were observed during the FLAME-4 study, suggesting denuder
losses were overcorrected by 10 % in some cases. Because this error is
larger than what would be expected based on Fig. 2, we used these in situ
observations to justify the application of a ±10 % uncertainty to
our denuder-derived absorption enhancements due to uncertainty in particle
losses (see Sect. 2.5 for details).
Comparison of absorption coefficients measured by dry and denuded
channels during thermal denuder bypasses for a day with significant
differences in the 405 nm channel. (a) 405 nm, (b) 660 nm.
Because absorption enhancements derived from thermally denuded channels in
this study were obtained by comparing denuded and dry channels at the same
wavelength run in parallel, it was critical to ensure that both the dry and
denuded channels at a given wavelength gave nearly identical absorption
coefficients when the denuder was not inline. To verify this, numerous
thermal denuder bypasses were completed with the purpose of ensuring that
enhanced absorption observed in dry channels was caused by coatings and not
by instrumental drift of other issues. During a thermal denuder bypass,
inlet air was diverted around the thermal denuder and directly to the cell
that normally measured denuded aerosol. Several thermal denuder bypasses
were done during each day and the absorption coefficients measured by the
dry and denuded channels were compared when measuring identical input
aerosol. Figure 3 below shows a comparison of absorption coefficients for
dry and denuded channels measured during thermal denuder bypasses. For
660 nm the slope between the absorption coefficients measured by the dry and
denuded channels is within 3 % of unity. However, at 405 nm, the two
channels differ by as much as 52 % (a day with one of the largest observed
differences is shown). This difference is largely due to the relatively
large (18 %, 1σ) error in the calibration of the 405 nm channel,
but is still at the outer limits of the Gaussian error curve created when
the errors are added in quadrature. The idea that the large differences
between the channels are more than random error is supported by the
observation that the dry channel was found to be consistently higher than
the denuded channel. The very high correlation (R above 0.99) between the
denuded and dry channels gives confidence that this is an error in the
absolute calibrations of the channels and that the magnitude of the
instrument response through the day is consistent between the two channels.
The reason for the denuded 405 nm channel having consistently lower signal
than the dry 405 nm channel remains unclear. However, during the last 3 days
of experimentation, new critical orifices were added to the ozone calibration
system allowing the introduction of higher levels of ozone. The calibration
of the dry 405 nm channel determined without the high ozone points (what was
done for most of the project) was consistently closer to the slope
determined using all ozone concentrations (including the high ozone points)
than the calibration of the denuded 405 nm channel without the high-ozone
points. These results suggest that the dry channel calibrations were more
accurate than the denuded channel calibrations, and because of this the
denuded channel absorption was adjusted to match the dry channel absorption.
The adjustment factors for each day of measurement are provided in Table S1 in the Supplement.
Absorption enhancement and absorption Ångström exponent
Absorption enhancement is defined here as the amount of absorption observed
when all material is present (BC and coatings) vs. the absorption observed
when just BC is present. One way to estimate absorption enhancement is to
measure the absorption of particles that have been thermally denuded at a
temperature high enough to remove organics vs. the absorption from
unperturbed particles (Lack et al., 2012a):
EAbsDen=babs_dry(λ)babs_den(λ),
where babs_dry(λ) is the absorption coefficient
of ambient particles at a specific wavelength and babs_den(λ) is
he absorption coefficient at the same wavelength after
the particles are heated in a thermal denuder. The thermal denuder in this
experiment was run at 250 ∘C. The error in the EAbsDen
calculated in this way does not depend on the absolute accuracy of either
the dry or the denuded absorption, because denuded absorption was adjusted to
match the dry absorption during thermal denuder bypasses, and while the
corrections can be large they were very stable throughout a given day
(Fig. 3). Instead, the dominant errors in this calculation are the
possibility that the denuder does not remove all organics and the need to
adjust for particle losses in the denuder. The minimum value of EAbsDen
calculated in this study was ∼ 0.91, observed during
measurements of a fire with a very high MCE and
presumably very little non-BC material being emitted. This result is only
possible if particle loss in the thermal denuder is overcorrected, and based
on this we derive the error in EAbsDen due to particle loss corrections
to be ±10 % and add this uncertainty in quadrature to 1 standard
deviation of the average measured value to report the total uncertainty in
the EAbsDen. This total error does not account for the potential error
of not fully removing organics in the denuder, since this error cannot be
quantified. AAE is defined as babs = aλ-AAE where babs
is the absorption coefficient and the constant, a, is independent of
wavelength. AAE is estimated from the slope of least-square fit of the
logarithm of absorption coefficients vs. the logarithm of wavelengths. AAE
are determined from the three wavelengths (405, 532, and 660 nm).
Modified combustion efficiency (MCE)
The MCE is defined as follows (Ward and Radke, 1993; Yokelson et al.,
1997):
MCE=ΔCO2ΔCO+ΔCO2,
where ΔCO and ΔCO2 are the mixing ratio enhancements
above background. Background mixing ratios were measured before the ignition
of each burn. The CO and CO2 mixing ratios were measured by an open-path Fourier transform infrared spectrometer (Stockwell et al., 2014). The
MCE reported in this study is the fire-integrated value.
Determination of elemental carbon to organic carbon ratio (EC / OC)
EC / OC estimates were made in identical fashion to Pokhrel et al. (2016), but
are described here again for clarity. Fine particulate matter (PM2.5) was
selected by a cyclone operating at a flow rate of 42 L min-1 and was
collected on to 37 mm quartz fiber filters (QFF; PALL, Port Washington, NY)
at ambient temperature. Field blanks were collected at a rate of one in
seven samples. Prior to use, QFF were pre-cleaned by baking at 550 ∘C for
18 h. Filters were stored in cleaned aluminum-foil-lined Petri dishes sealed
with Teflon tape, and stored frozen (-20 ∘C) before and after analysis. OC
and EC were measured by thermal optical analysis (Sunset Laboratories,
Forest Grove, OR, USA) following the IMPROVE-A protocol where the EC / OC
split was determined by thermal optical transmittance. The effects of
positive sampling artifacts due to carbonaceous gas adsorption were assessed
using quartz filters behind Teflon (QBT; Cheng et al., 2009) for 14 of the
96 fires, including grass, rice straw, ponderosa pine, black spruce, and
peat. For fires with QBT collected, the OC on the backup filter was
subtracted directly. For fires without backup filters or those that were
below the detection limit, the average OC correction for that fuel type was
applied: rice straw (2.0 ± 0.4 %), ponderosa pine (1.2 %), black
spruce (2.9 ± 1.6 %), and peat (3.1 ± 0.8 %). For fuel
types without backup filters collected, the study average OC artifact
(2.4 ± 1.2 %) was subtracted. This approach to artifact correction
assumes that the amount of carbonaceous gas adsorbed is proportional to the
mass concentration of OC; this assumption is considered to be reasonable
because the back-up filters contained less than 5.6 µg OC cm-2
and similar quartz fiber filters become saturated above 6 µg OC cm-2
(Turpin et al., 1994). Analytical uncertainties for OC were
propagated from the standard deviation of field blanks (0.7 µg cm-2)
and 5 % of the OC concentration. For EC, uncertainties were
propagated from an estimate of the instrument precision (0.1 µg m-2),
5 % of EC concentration and 5 % of pyrolyzed carbon (which
forms from OC charring on the filter during analysis). The value of 5 %
is a conservative estimate of the precision error in replicate sample
analysis, which is typically 1–3 % (NIOSH, 1999). Analytical
uncertainties for the EC / OC ratio were propagated from the individual EC
and OC uncertainties.
Results and discussion
Absorption enhancement derived with a thermal denuder
Absorption enhancement determined by comparing aerosol passed through a
thermal denuder at 250 ∘C to non-thermally-denuded aerosol
(EAbsDen, Eq. 1) was calculated during 22 individual burns of 12
different fuels. The EAbsDen values reported in this study are the
average value obtained over approximately 1 h of measurements that were
made after the smoke had completely mixed in the combustion room. Figure 4
shows bar plots of the average EAbsDen for different fuels. Repeated
burns of the same fuel often generated different burn conditions resulting
in different MCEs (Eq. 2), which are also given. The total bar height (red
plus blue) represents the absorption enhancement at 405 nm and red bars
represent EabsDen at 660 nm. Fuels are categorized into four different
groups: coniferous trees, crop residues, grasses and brushwood, and peat.
These fuel categories have large contributions to the total biomass burning
across different parts of world (Page et al., 2002; Chang and Song, 2010;
Clinton et al., 2006; McCarty et al., 2007). For all burns, EAbsDen
is larger at 405 nm than at 660 nm, except for a giant saw grass (GSG)
burn where the two are nearly identical. In this burn, both the 405 and
660 nm EAbsDen are unity within experimental uncertainty, suggesting that the BC
emitted during this burn had very little coating. The fact that all burns,
except giant saw grass and wire grass, have an EAbsDen at 405 nm
significantly larger than that at 660 nm provides evidence of the presence
of BrC on most of the burns since the lensing effect typically has a weak
dependence on wavelength (Lack and Cappa, 2010; McMeeking et al., 2014).
Bar plots of absorption enhancements derived by thermally denuding
aerosol and measuring denuded and non-denuded particles at both 405 and 660 nm.
Results are grouped in terms of fuel types. Total bar heights (red + blue)
are representative of absorption enhancement at 405 nm while red bars
represent absorption enhancement at 660 nm. Panel (a) is for coniferous
trees, (b) is for crop residues, (c) is for grass and
brushwood, and (d) is for peat. The legend shows the name of each fuels reported
in each group.
EAbsDen at 660 nm ranges from 0.92 ± 0.09 to 1.43 ± 0.17,
similar to the range of EAbsDen observed during FLAME-3 for a different
suite of fuels. FLAME-3 EAbsDen results at 532 and 781 nm ranged
from 1.2 to 1.5 (McMeeking et al., 2014) but the thermal denuder in that study was
operated at or below 150 ∘C while the thermal denuder in this study was
run at 250 ∘C. Our results at 660 nm are also similar to the EAbs
of 1.4 measured at 532 nm in a biomass plume near Boulder, CO, USA (Lack et al.,
2012a). The range of EAbsDen at 405 nm observed in this study is
similar to previous studies (McMeeking et al., 2014; Lack et al., 2012a),
except for peat emissions where much higher absorption enhancement is observed. The
peat burns give a very high value (5.65 ± 1.43) of EAbsDen at
405 nm because smoldering emissions from peat are predominantly BrC with a
negligible amount of BC content (Chakrabarty et al., 2016; Pokhrel et al.,
2016; Stockwell et al., 2016). It is evident from Fig. 4 that EAbsDen
can vary significantly, depending on burn conditions, even for the same fuel.
405 nm absorption enhancement parameterized with (a) AAE,
(b) EC / OC, and (c) MCE. The symbols are sorted by fuel type and listed in the
legend. Red lines are least-square fits of the data with the equation and
correlation coefficient (r) reported for each case. Given the near absence
of BC, peat burns are considered as outliers and not included in model fit.
Parameterization of absorption enhancement
In order to gain a better understanding of what drives variation in
EAbsDen, we examined correlations between absorption enhancement and
other fire-relevant variables. It is notable that some regressions are done
for a semi-log plot while others are linear or log–log. The type of
regression was chosen based on objective criterion for simple regression.
Namely that the residuals are equally scattered from the regression line and
that the residuals are as close as possible to a normal distribution. The
model (either logY vs. logX, logY vs. X, or Y vs. logX) which satisfied these
criterion for simple linear regression was chosen. Fuel type alone is
insufficient because EAbsDen varies dramatically during different burns
of a single fuel. Figure 5 shows EAbsDen at 405 nm vs. the
absorption Ångström exponent, elemental to organic carbon ratio (EC / OC), and
MCE (Eq. 2). AAE values in Fig. 5 were
calculated from a best fit to the logarithm of absorption coefficient
vs. wavelength at three wavelengths (660, 532, 405 nm), as detailed in
Pokhrel et al. (2016). A strong positive correlation (r = 0.96) between
absorption enhancement at 405 nm and AAE is observed for all fuels. The
linear relationship between logarithm of EAbsDen and AAE suggests that
BrC absorption can be parameterized within the AAE framework. While
the exact nature of the trend is notable, the trend itself is expected
because the AAE of pure BC is typically near 1 (Kirchstetter and Thatcher,
2012; Wiegand et al., 2014) and anything larger than that strongly suggests
the presence of absorbing coatings.
Same as Fig. 5 but for 660 nm.
Figure 5a shows that for AAE less than 2, EAbsDen at 405 nm remains
close to 1, indicating little influence from BrC or coatings when emissions
have a low AAE. Also for AAE less than 2, EAbsDen at 660 nm
(Fig. 6a) are similar to EAbsDen at 405 nm, again indicating little
presence of BrC, which is expected to absorb mainly at shorter
wavelengths. Although EAbsDen at 405 nm shows strong correlation with
AAE, EAbsDen at 660 nm does not show good correlation, with a
Pearson's correlation coefficient for logarithm of EAbsDen vs. AAE of
0.96 at 405 nm but only 0.32 at 660 nm. This result strongly suggests that
either BrC is not present at 660 nm or, if there is BrC at
660 nm, it is not strongly correlated to AAE. This means that, if
EAbsDen 660 nm is purely from lensing, the effect of non-absorbing
coatings on absorption cannot be easily parameterized with AAE.
Figures 4b and 5b show that EAbsDen at both 405 and 660 nm
decrease linearly (r = -0.89 and -0.78 respectively) with the logarithm of
the EC / OC ratio. When aerosol composition has more EC than OC, EAbsDen
at 405 and 660 nm both approach 1 as the effects of lensing and BrC become
minimal. The slope of the 660 and 405 nm EabsDen vs. EC / OC fits are
very different and as the fraction of OC increases, EAbsDen at 405 nm
grows much more quickly than EAbsDen at 660 nm. Saleh et al. (2014)
estimate that the average BC core and particle size during FLAME-4 was
100 and 200 nm respectively. In these size ranges, core-shell Mie theory
predicts that the lensing effect will be similar at 660 and 405 nm
(McMeeking et al., 2014). Given this, the much larger EabsDen at 405 nm
indicates absorption from BrC. A key observation is that EAbsDen at
405 nm can be parameterized with EC / OC without the need to explicitly define
fuel type. There is relatively poor correlation between EAbsDen at
either 405 or 660 nm with MCE. There are fewer data points for EC / OC
parameterizations because not all burns measured with the optical suite had
corresponding EC / OC measurements. Accordingly, the linear fit of
EAbsDen vs. MCE at 405 nm does not include peat and one ponderosa pine
burn (with MCE ∼ 0.83) because there was no EC / OC data for
these burns and they represented significantly lower MCE than all other
burns. For similar MCE values, EAbsDen at 405 nm varies by a factor
of 3 in some burns. A potential reason for the poor fit with MCE is the
difficulty of MCE to predict aerosol properties such as BC / OA or EC / OC
(Grieshop et al., 2009; Pokhrel et al., 2016) on which absorptivity of
organic aerosol has a strong dependence (Saleh et al., 2014).
Contribution to total absorption from brown carbon and lensing
The dataset collected during this study allows us to estimate the
contribution due to BC, BrC, and lensing in several different, commonly
implemented, ways as discussed in the introduction. Here we describe the
specific approaches used and evaluate the difference in the results
generated by these approaches. A conceptual representation of these
approaches is shown in Fig. 7.
Conceptual representation of the three different approaches used
to estimate the fraction of total absorption contributed by BC, BrC, and
lensing. Numbers in parentheses correspond to the approach represented by
that line. The y axis is normalized so that the denuded absorption of the
660 nm channel is unity.
Approach 1: assume the thermal denuder removes all organic carbon
and assume EAbs from lensing is constant at all wavelengths
In this approach EAbsDen at 660 nm is assumed as an enhancement due
to lensing. We assume the absorption enhancement from lensing is identical
at 405 and 660 nm (Guo et al., 2014; Nakayama et al., 2014; Cappa et al.,
2012). We assume the absorption of the denuded channel at 405 nm is due
entirely to BC. The remaining EAbsDen after lensing is subtracted is
assumed to be caused by BrC. The following equations summarize
these assumptions and describe how we derive the absorption from BrC at 405 nm.
babs_405_BrC=babs_405_dry-Eabs_660×babs_405_den,
where babs_405_BrC is absorption due to
BrC at 405 nm, babs_405_dry is
non-denuded dry absorption measured at 405 nm, babs_405_den
is denuded absorption measured at 405 nm, and
Eabs_660 is absorption enhancement at 660 nm. Due to
incomplete removal of low-volatile organics, estimated fraction of
absorption due to BrC using approach 1 is most likely underestimated. The
logic is that Eq. (3) can be simplified as
babs_405_BrC=babs_405_dry-Eabs_660×babs_405_denbabs_405_BrC=babs_405_dry-babs_660_drybabs_660_den×babs_405_den.
The denuded absorption at both 660 and 405 nm will be overestimated due to
incomplete removal of organics, but the problem is expected to be worse at
405 nm because both BrC and lensing increase the 405 nm denuded
absorption while lensing is the dominant effect for the 660 denuded
absorption. Given this, the ratio babs_405_denbabs_660_den is
expected to be larger than 1 and hence BrC absorption will be
underestimated because neither of the dry absorptions (405, 660 nm) will be
affected. To extend our results to 532 nm, where dry, but not denuded,
absorption was measured by the PAS, we calculated the absorption due to BrC as follows:
babs_532_Brc=babs_532_dry-Eabs_660×babs_405_den×405532AAEden,
where babs_532_Brc is absorption due to
BrC at 532 nm, babs_532_dry is absorption
measured at dry phase at 532 nm, Eabs_660 is absorption
enhancement at 660 nm, and AAEden is the absorption Ångström exponent
calculated between the 660 and 405 nm denuded channels. The BrC
absorption estimated by Eqs. (3) and (4) can be visualized by subtracting
dry absorption and the line labeled “BC + Lens (1)” in Fig. 7.
Approach 2: assume 660 nm denuded absorption represents BC
absorption, assume the AAE of BC is 1, and assume EAbs from
lensing is constant at all wavelengths
In this approach we assume the absorption measured in the denuded 660 nm
channel is due entirely to BC. We estimate absorption due to BC at
wavelengths less than 660 nm using an AAE of 1 for BC. This approach may be
more accurate than approach 1 if the thermal denuder does not effectively
remove all of the BrC absorption at 405 nm. This approach will be
incorrect if the AAE of BC is different than 1, if there is BrC at
660 nm that is not removed by the denuder or if the denuder generates BC at
660 nm. EAbsDen at 660 nm is considered to be due to lensing of the BC
core and the same lensing effect is applied on all wavelengths. Absorption due
to BC at any wavelength is estimated by using Eq. (5) (line “BC (2,3)” in
Fig. 7),
bmabs_λ1_BC=babs_660_den×660λ11,
and absorption due to BrC is estimated by the following:
babs_λ1_BrC=babs_λ1_dry-Eabs_660×babs_λ1_BC.
Substituting the value of babs_λ1_BC from
Eq. (5) generates an alternate expression for absorption due to BrC:
babs_λ1_BrC=babs_λ1_dry-Eabs_660×babs_660_den×660λ11,
where λ1 is the desired wavelength, which in this study is
405 and 532 nm. Absorption from lensing is calculated as the difference between
total absorption and the sum of the contributions from BC and BrC to
absorption. BrC absorption estimation using Eq. (7) can be visualized by
subtracting line “BC + Lens (2)” from the dry absorption in Fig. 7.
Alternate description of approach 2
A widely used approach is to assume that absorption at 660 nm or higher
wavelengths is equivalent to BC absorption plus lensing from the clear
coating because there is no absorption contribution from BrC. It is also
assumed that coated BC particles have an AAE of 1, similar to uncoated BC.
Absorption due to BC at lower wavelengths is estimated as
babs_λ1_BC=babs_660_dry×660λ11
and absorption due to BrC is estimated as
babs_λ1_BrC=babs_λ1_dry-babs_660_dry×660λ11.
Interestingly, because babs_660_dry = Eabs_660 babs_660_den, Eqs. (9) and (7) are equivalent. Therefore, this
approach gives the same absorption due to BrC as what was described for
approach 2 and there is no need to call this a separate approach. It is
described here because it is widely implemented by groups without a thermal
denuder and to demonstrate that several different assumptions lead to the
same numerical result for BrC absorption. Importantly, Eqs. (8) and (9)
do not allow for assessment of the impact of lensing while Eqs. (5)
and (6) do. A final note is that one also arrives at Eq. (9) if it is
assumed that lensing has a negligible impact on BC properties and all
absorption at 660 nm is from BC. However, this is a rather unrealistic
approach given that lensing has clearly been shown to impact absorption at 660 nm.
Approach 3: assume 660 nm denuded is BC absorption, assume 660 nm
dry is BC plus lensing, assume clear-coated BC has AAE of 1.6
AAE for non-denuded BC has been commonly assigned as 1.0 in many past
studies, including some recent studies (Kirchstetter and Thatcher, 2012;
Wiegand et al., 2014). However, it has been shown theoretically that with a
non-absorbing coating, AAE of BC can be as large as 1.6 (Gyawali et al.,
2009; Lack and Cappa, 2010). Here we use AAE of 1 for uncoated BC and AAE of 1.6
for coated BC. This approach estimates an approximate maximum increase
in lensing with decreasing wavelength rather than assuming it is constant
with wavelength as in the other approaches. It is an important reference
point because while BrC may bleach (Wang et al., 2016; Forrister et al.,
2015; Lee et al., 2014), absorption enhancement from lensing will remain
unless coatings evaporate. Absorption due to BC, lensing, and BrC is calculated as follows:
Absorption due to BC is estimated as
babs_λ1_BC=babs_660_den×660λ11
and absorption due to BrC is estimated by
babs_λ1_BrC=babs_λ1_dry-babs_660_dry×660λ11.6,
where babs_λ1_BC is
absorption due to BC at wavelength λ1, babs_660_den is denuded absorption measured at 660 nm,
babs_λ1_BrC is absorption due
to BrC at λ1, babs_λ1_dry is non-denuded absorption measured at
λ1, babs_660_dry is
non-denuded absorption measured at 660 nm, and λ1 is the
desired wavelength (405 and 532 nm in this study). BrC estimated by
Eq. (11) can be visualized as the difference between the dry absorption and line
“BC + Lens (3)” in Fig. 7. Absorption contribution due to lensing is
estimated by subtracting absorption due to BC and BrC from the total absorption.
Percentage of absorption due to BC, lensing (clear coating), and
BrC from biomass burning aerosol emissions at 405 nm estimated from three
different approaches. The ID is the fire ID assigned during FLAME-4 for
particular burns. The ratio in the rightmost column is the ratio in BrC
absorption estimated by approach 2 vs. approach 1. Results where the BrC
contribution was found to be zero by approach 1 are not assigned a ratio. NA
indicates not available.
ID
Materials
Approach 1
Approach 2
Approach 3
BC
Coat
BrC
BC
Coat
BrC
BC
Coat
BrC
Ratio
129
Pine
38
3
58
14
1
85
14
6
80
1.5
142
Pine
68
11
21
48
10
42
48
30
22
2.0
144
Pine
55
21
23
30
14
57
30
29
42
2.5
130
California rice straw
64
2
34
37
2
62
37
15
49
1.8
143
California rice straw
35
4
61
16
2
82
16
8
76
1.3
131
Black Spruce
59
2
39
37
2
60
37
16
46
1.5
134
Black Spruce
55
10
35
35
6
59
35
20
45
1.7
138
Organic Hay
NA
NA
NA
29
11
60
29
24
47
NA
146
Organic Hay
41
7
53
28
0
72
29
9
62
1.4
132
Organic Wheat
70
1
29
48
1
52
48
17
35
1.8
149
Organic Wheat
73
11
16
40
6
54
40
21
38
3.4
139
Giant saw grass
100
0
0
86
0
14
86
14
0
-
148
Giant saw grass
NA
NA
NA
NA
NA
NA
66
22
12
NA
133
Conventional Wheat
68
1
32
45
1
54
45
17
38
1.7
135
Chamise
93
0
7
64
0
37
64
21
15
5.3
136
Manzanita
92
0
8
73
0
28
73
24
3
3.5
141
Wire grass
100
0
0
100
0
0
100
0
0
–
147
Sugar cane
40
5
55
21
2
77
21
10
69
1.4
150
NC peat
16
2
82
8
0
92
8
3
89
1.1
Based on these approaches, the contribution of BC, BrC, and lensing to total
absorption by biomass burning aerosol under different burn conditions is
estimated. Table 1 summarizes the results at 405 nm. BrC at 405 nm is
estimated to contribute 0 to 92 % of total biomass burning aerosol
absorption depending upon the burn and the approach used. From multiple
burns of the same fuel (pine, California rice straw, black spruce, organic hay,
organic wheat) it is evident that the BrC contribution to absorption changes
significantly for a single fuel demonstrating the importance of burn
conditions. Between the three different approaches, the lower bound for the
contribution of BrC estimation is consistently approach 1 and the upper
bound is consistently approach 2, with approach 3 being in the middle. We
hypothesize that the reason for approach 1 consistently resulting in the
lowest fraction of absorption from BrC is incomplete removal of organics by
the thermal denuder, an idea that is supported by the observation that the
AAE of the denuded aerosol channels (405 and 660 nm) was often significantly
larger than one, which is often observed for uncoated BC (average
AAEden = 1.9). It is thought that the resonance time of aerosol in
the thermal denuder may have been insufficient to remove the significant
coatings present in some burns or that the temperature (250 ∘C) could
have been too low to remove extremely low-volatility organic compounds that
have been suggested to be important contributors to BrC (Saleh et
al., 2014). While this suggests that the BrC contributions derived from
approach 2 may be closer to reality, approach 1 provides a useful lower
bound to BrC absorption and even this lower bound is often a significant
fraction of total absorption. The difference in BrC contribution predicted
by approach 1 and 2 varies from burn to burn with a maximum ratio of 4.3 and
a mean ratio of 2.1, demonstrating the significant variation between
methodologies and the potential difficulties of assessing BrC via thermal
denuding. Similar to the 405 nm results, the BrC contribution at 532 nm is
important to overall absorption and shows large variations, ranging from
0 to 58 % as shown in Table S2. The percentage of absorption due to BrC
at 532 nm also has a strong dependency on fuel type and the methodology
used, similar to the 405 nm results.
The maximum ratio of the BrC contributions found in approach 1 and 2 is
2.5 at 532 nm, lower than the maximum difference at 405 nm. However, the mean
ratio is 2.1, which is identical to the 405 nm results. We find that
coatings can contribute up to roughly 30 % of the absorption, but
generally the contribution due to coating is low relative to BC and BrC.
However, it should be noted that incomplete removal of organics by the
thermal denuder would result in an underestimate of absorption enhancement
from clear coatings (lensing) and an overestimate of the relative importance
of BC in all three approaches. Approach 3 consistently yields the highest
contribution from coatings, suggesting that an Ångström exponent of 1.6 for
BC with clear coatings is indeed at the high end of possible AAE values.
Percentage of absorption due to BrC vs. EC / OC – (a) for
approach 1, (b) for approach 2, and (c) for approach 3. Blue symbols are
for 405 nm and green symbols are for 532 nm. Panels (d)–(f) are
the same as the top row of data except vs. AAE.
Our estimation of BrC contribution to total biomass burning aerosol
absorption shows large variations, but it is also clear that, independent of
methodology, the contribution of BrC to total absorption is large and often
of similar or greater magnitude than the contribution from BC. Most ambient
studies show relatively low (< 20 %) BrC contribution to
absorption at 405 nm (Yuan et al., 2016; Nakayama et al., 2014; Cappa et
al., 2012), but these studies were not focused on biomass burning emissions
and had significantly less biomass burning influence. Lack et al. (2012a)
estimated that BrC contributed 27 ± 15 % of absorption at 405 nm
during a wild fire in Boulder, Colorado, based on thermal denuder
measurements. Additional studies focused on biomass burning reported ranges
of values for the contribution to absorption from BrC (McMeeking et al.,
2014; Flowers et al., 2009) and these ranges are similar to ours. Flowers et
al. (2009) showed BrC can contribute 27–51 % of absorption at 405 nm using
literature mass absorption cross section (MAC) values to estimate the
absorption due to BC and calculating absorption due to BrC by comparing
measured absorption with estimated absorption based on MAC and BC mass.
McMeeking et al. (2014) found, based on fresh emissions from a wide range of
biomass fuels, that non-refractory particulate matter can contribute
20–80 % of the light absorption at wavelengths ≤ 532 nm, a result
similar to the range of contributions to absorption from BrC
(0–92 %) found in this study.
Parameterization of brown carbon absorption
Given the large variation in the contribution of BrC to total absorption for
different fuels and for repeated burns of the same fuel, it is clear that
some type of framework for estimating the significance of BrC
absorption is needed. Saleh et al. (2014) developed a methodology to
estimate the imaginary part of the refractive index of BrC based on
the ratio of black carbon to organic aerosol mass. This approach requires
detailed particle size measurements that were not available in the current
study. Instead, we directly parameterize the contribution of BrC to biomass
burning aerosol absorption with MCE, EC / OC, and AAE. Figure 8 shows the
percentage of absorption due to BrC at 405 and 532 nm estimated with
approaches 1–3 vs. EC / OC and AAE. The general pattern is that the
fraction of absorption due to BrC is highest when the aerosol composition is
dominated by OC. On the other hand, aerosol composition dominated by EC shows
a lower contribution to total absorption from BrC at both wavelengths.
Generally, the trend of the fraction of BrC absorption vs. EC / OC is similar
for different approaches, but the slopes of the lines change significantly
based on the approach. The percentage of absorption due to BrC at 405 and
532 nm shows a reasonably good correlation with EC / OC, as demonstrated by the
correlation coefficients between 0.7 and 0.9 shown in Fig. 8a–c.
We also parameterize the BrC percentage with AAE and find that AAE
shows very good correlation with the percentage of absorption due to BrC
yielding correlation coefficients between 0.8 and 0.99. Figure 8d–f
shows that for an AAE less than 2, contributions from BrC are less than
about 20 % and decrease with decreasing AAE regardless of the approach
used to estimate these values. Also, there is a higher contribution to
absorption from BrC at 405 nm vs. 532 nm at AAEs above roughly 2.5. On the other
hand, MCE (Fig. S1 in the Supplement) does not have an easily fitted relationship with the
percentage of absorption from BrC at either wavelength. For an MCE of 0.92
or greater, the percentage of absorption by BrC varies by a factor of 3 to 4
for very similar MCE values. The coefficients for each fitted line in Fig. 8
are reported in Table S3.