Introduction
Black carbon aerosols in the atmosphere play a vital role in climate change
by absorbing solar radiation and altering the formation, lifespan and albedo
of clouds (Novakov et al., 2005; Ramanathan and Carmichael, 2008; Bond et
al., 2013). It was operationally defined according to its light absorption
capacity, chemical reactivity and/or thermal stability (Lack et al., 2014).
One definition is refractory black carbon (rBC), which corresponds to the
carbon mass derived from laser-induced incandescence (LII) emission at a
boiling point at 4000 K. Open biomass burning (which normally refers to
burning of living or dead vegetation such as agriculture residue, grass,
forest, leaves, and shrub due to anthropogenic activities and natural causes,
abbreviated as OBB) is one of the important sources of rBC, and it
contributes ∼ 42 % of atmospheric loadings in the global emissions
budget (Bond et al., 2004). In addition to rBC, OBB also simultaneously emits
substantial amounts of semi-volatile organics that undergo extremely
complicated mixing processes with rBC during transport. Jacobson (2001)
pointed out that the light absorption by internally coated rBC by
inorganic/organic matter could increase, due to the “lensing effect”, in
which a non-absorbing coating directs more light to the cores of rBC
particles. However, the debate on the absorption enhancement capacity of
rBC-containing particles is still ongoing, because discrepancies exist
between observations and theoretical predictions based on Mie scattering
models (Shiraiwa et al., 2010; Cappa et al., 2012) and among observation
results from different locations and using different sources (Healy et al.,
2015; S. Liu et al., 2015; Massoli et al., 2015; Ueda et al., 2016). The
widely accepted explanations are as follows. First, the morphologies of rBC
particles differ among different sources, and the process of particle aging
in the atmosphere changes the physical structure of the particles. For
instance, several studies have found that rBC particles with a fractal
structure tend to collapse to a more closely packed shape when they are
thickly coated (Adachi et al., 2010; Chen et al., 2010; He et al., 2015).
Modeling study indicated that hygroscopic growth of rBC-containing particles
also results in more compact rBC cores (Fan et al., 2016). Second, the rBC
particles sometimes were also reported to be attached on the surface of
non-rBC matter (Moteki et al., 2014). Either of these circumstances renders
the core–shell model invalid or introduces biases into the results. Sedlacek
et al. (2012) reported that a large fraction (60 %) of rBC-containing
particles with non-core–shell structures exist in biomass burning plumes.
Methodologies have also been developed to distinguish particles with attached
rBC (bare rBC on the surfaces of non-rBC particles) from rBC-containing
particles with the core–shell structure (Moteki et al., 2014). In addition,
OBB is an important source of brown carbon (BrC), which has distinct light
absorbing features with different wavelength dependence, and their
coexistence of rBC and BrC also influences the overall absorption enhancement
of rBC-containing particles (Lack et al., 2012; Saleh et al., 2014; D. Liu et
al., 2015; S. Liu et al., 2015; Saleh et al., 2015).
Biomass normally consists of cellulose/hemicellulose, organics and water. The
emissions of rBC from OBB are determined by the composition (carbon content)
of the biofuel and the evolution of combustion (Yokelson et al., 1997;
Andreae and Merlet, 2001). Briefly, the combustion process of biomass begins
with the pyrolysis of biofuel molecules and the evaporation of flammable
mixtures (i.e., volatile compounds) and water, which is followed by flaming
combustion that converts most carbon substances to carbon dioxide (CO2).
rBC particles are also produced in large quantities at this stage due to the
oxygen-limited conditions and high temperatures. The last stage is smoldering
combustion, which predominantly emits carbon monoxide (CO) and organics
(Andreae and Merlet, 2001). It was reported that “tar balls” that mostly
consisted of BrC and secondary organic aerosols with low volatility were also
emitted during smoldering combustion (Pósfai et al., 2004). Detailed
descriptions of the physical properties of biomass burning particles have
been provided in the literature (Reid et al., 2005; Akagi et al., 2011). The
modified combustion efficiency (MCE), which is defined as ΔCO2/(ΔCO2+ΔCO) (Δ is the difference in
measured concentrations between the OBB plume and the corresponding
background value), has been used to indicate the relative amounts of flaming
and smoldering combustion during a fire for characterizing the emission of
rBC and organic matter. An MCE value > 0.95 is normally regarded
as flaming-dominant combustion, whereas MCE < 0.9 represents
smoldering-dominant combustion (Kondo et al., 2011; Pan et al., 2012, 2013;
May et al., 2014).
The rBC emission ratio (ΔrBC / ΔCO), which is defined as
the enhancement of mass concentration rBC (in unit ng m-3) with respect
to its background versus that of CO (in unit ppbv, parts per billion volume),
is an applicable indicator for constraining the rBC emission inventory for
models (Pan et al., 2011). The variability in ΔrBC / ΔCO
among observational studies mostly results from differences in measurement
techniques, fuel type, and burning conditions. For example, observations made
onboard the NOAA WP-3D aircraft yielded ΔrBC / ΔCO values
of 9 ± 2 ng m-3 ppbv-1 (Spackman et al., 2008) and
17.4 ng m-3 ppbv-1 (Schwarz et al., 2008) for brush fire plumes
during the TexAQS field campaign. Airborne observations on the NASA DC-8
aircraft indicated that the ΔrBC / ΔCO values were
8.5 ± 5.4 ng m-3 ppbv-1 for plume of boreal forest and
agriculture fires in Asia, and 2.3 ± 2.2 ng m-3 ppbv-1 for
wildfire plume in North America (Kondo et al., 2011). Observations using a
multi-angle absorption photometer (MAAP, which employs the filter-based light
absorption technique; here we consider BC instead of rBC) at mountain sites
(30.16∘ N, 118.26∘ E; 1840 m a.s.l.) in South China
yielded high ΔBC / ΔCO values
(10–14 ng m-3 ppbv-1) when the site was subjected to burning of
crop residues (Pan et al., 2011). The large variability also resulted from a
difference in sampling altitude. For instance, airborne-based measurements
tend to capture flaming-dominant plumes because they are more easily injected
to high altitudes than smoldering-dominant plumes (Kondo et al., 2011). In
fact, rBC emissions are heavily dependent on the combustion state of biomass.
Field measurements using a semi-continuous ECOC analyzer (thermo-optical
transmittance technique, IMPROVE protocol) during OBB episodes in East China
yielded ΔEC / ΔCO values of
17.4 ± 5.2 ng m-3 ppbv-1 for flaming-dominant cases and
11.8 ± 2.3 ng m-3 ppbv-1 for smoldering-dominant cases
(Pan et al., 2012). Biomass burning experiments in the laboratory on
15 individual plant species sampled in the United States indicated that
ΔrBC / ΔCO increased by up to
40 ng m-3 ppbv-1 as MCE increased to 0.95, and this result was
largely insensitive to the biomass type used (May et al., 2014). Note that
ΔrBC / ΔCO values decreased significantly with aging of
OBB plumes, owing mostly to the below-cloud and in-cloud scavenging processes
as a result of large fractions of water-soluble organic species (Mazzoleni et
al., 2007; Gilardoni et al., 2016). Observations made in both East Asia and
North America indicate a strong dependence of ΔrBC / ΔCO
on accumulated precipitation along backward trajectories (Kondo et al.,
2011), implying that the rBC-containing particles became hydrophilic and were
removed by wet deposition during transport. This result is consistent with
observations of OBB plumes at the top of Mt. Tai (36.26∘ N,
117.11∘ E; 1534 m a.s.l.) in North China, which indicated that
ΔBC / ΔCO values for OBB plumes decreased substantially
due to cloud scavenging processes (Pan et al., 2013).
The Yangtze River Delta region (YRDR) is one of the most important
agricultural regions in China, and it accounts for 29 % of total grain
production. Wheat and rapeseed plants are two major crops. After harvest,
some of the crop straws are burned in the open air at fields, resulting in
severe air pollution on the regional scale. Although field measurements of
variations in the concentrations of rBC and CO and their ratios in OBB plumes
have been reported, their physical characteristics may change significantly
due to the rapid aging/mixing processes of semi-volatile organic vapors.
Therefore, laboratory studies are very important to obtain insight into the
initial emission features of rBC particles in China. In the present study, we
conducted laboratory burning experiments using two crop residues (wheat
straws and rapeseed plants) obtained from the YRDR. The mass concentration
and size distribution of rBC particles in the OBB plume were measured using a
single particle soot photometer (SP2). The physical properties of nascent
rBC-containing particles, the evolution of the size distribution, the mixing
state of the rBC particles, and their dependence on the combustion state were
investigated. The information presented in this paper is helpful for
constraining/reducing uncertainties in OBB emission inventories and the
estimates of their climatic effects by models.
Experiments
Description of the burning experiments
We conducted burning experiments in the laboratory using samples of wheat
straw and rapeseed plants that were collected in an agricultural area of East
China during a field campaign in 2010 (Pan et al., 2012). All of the biomass
was stored in sealed plastic bags to preserve its original state. During
experiments, the biomass sample was placed on an aluminum foil net rack in a
heat-resistant combustion box with an approximate volume of 144 L, and it
was ignited by a butane-fuel lighter from the bottom. Afterwards, the biomass
burned until it went out completely. In general, a total of 24 samples (the
mass of each sample was ∼ 20 g) were tested. The OBB smoke was removed
from the room by a venting fan through a flexible rubber hose at a flow rate
of 120 m3 h-1. The wheat straw samples were classified into two
groups. Twenty samples were burned in the chamber without artificial
treatments, and four of the samples were placed in humid conditions
(RH > 99 %) for 30 min to absorb moisture for comparison.
Detailed information on the setup of the OBB experiments is provided in the
literature (Inomata et al., 2015). As mentioned, MCE is a useful metric for
describing the combustion phase of biomass burning, and the calculation of
MCE requires simultaneous measurements of CO and CO2 concentration.
Here, the mixing ratio of CO2 was measured using a Li-7000 CO2
analyzer (Li-COR Inc.; detection range 0–3000 ppmv, RMS noise 35 ppbv,
integration time 0.5 s) through a separate 1/8 inch, ∼ 1.5 m long
Teflon tube. The mixing ratio of CO was concurrently measured with an
ultrafast CO analyzer (model AL5002, Aero-Laser GmbH; detection range
0–100 ppmv, detection limit 1.5 ppbv, integration time 1 s). Comparison
with a non-dispersive infrared CO gas analyzer (Thermo 48C; precision
10.0 ppbv, RMS noise 5.0 ppbv, average time 1 min) indicated that the
measurement uncertainty was within 5 %.
To avoid instrument overloading due to the extremely high concentrations of
particles and trace gases, the inlets of the sampling lines were situated
∼ 40 cm away from the combustion smoke. The sampling flows were
subsequently diluted in a dilution system by mixing dry zero air produced by
a zero gas generator (Thermo Inc., model 111). The flow rate of injected zero
air was precisely controlled with a mass flow controller (Kofloc Inc.,
model 3660; accuracy ±1.0 % at 25 ∘C). The uncertainty of
the dilution system was evaluated using known-size dry polystyrene sphere
latex particles (PSL, size standard particles, JSR Corporation, Japan). The
PSL aerosols were produced by a nebulizer at a flow rate of
3.5 L min-1, passed through a diffusion dryer (model 3062, TSI Inc.,
USA) and then size-selected using a differential mobility analyzer (DMA,
model 3081, TSI Inc., USA). The number concentration of PSL particles was
measured using a laser aerosol spectrometer (LAS-X II, PMS(GB) Ltd., UK;
uncertainty 5 %, flow rate 50 ccm) with and without dilution. The errors
of the dilution system were found to be 6, 2, 2, 4, and 5 % for particles
with mobility diameters of 120, 200, 300, 500, and 1000 nm, respectively, at
a dilution ratio of ∼ 50. During the burning experiment, a 50 cm long,
1/4 inch flexible conductive silicone tube (TSI Inc., USA) and stainless
steel Swagelok fittings were used for the aerosol tubing (dilution factor:
46), and a polytetrafluoroethylene (PTFE) tube and fittings were used for the
measurement of gases (dilution factor: 22). Abbreviations and symbols used in
this paper are shown in Table 1.
Abbreviations and symbols used in this paper.
Symbol/
Full name/explanation
abbreviation
rBC
Refractory black carbon, as derived using the LII method at a temperature of ∼ 4000 K.
Δt
The delay in the time of occurrence of the incandescence peak after that of the peak of the scattering signal
APM
Aerosol particle mass analyzer (Kanomax Inc.)
S/C ratio
Shell / core ratio
CO
Carbon monoxide
CO2
Carbon dioxide
CS
Scattering cross section
DMA
Differential mobility analyzer (TSI Inc.)
FS
Fullerene soot (C60)
LEO fitting
Leading-edge-only fitting method proposed by Gao et al. (2007)
LSP
Light scattering particle
MCE
Modified combustion efficiency
MED
Mass equivalent diameter
MMD
Mass mode diameter
non-rBC
Non-refractory black carbon matter that evaporates as rBC absorbs energy
OBB
Open biomass burning
SP2
Single particle soot photometer (DMT Technologies)
Instruments
Single particle soot photometer
A single-particle soot photometer (SP2, Droplet Measurement Technologies
Inc.) was used to examine the evolution of the number concentration and
mixing state of the OBB particles. The SP2 employs a continuous intra-cavity
Nd:YAG laser beam (1064 nm, TEM00 mode, Gaussian) to produce a strong laser
power field and detects the laser-induced incandescence signal emitted from
individual rBC particles when they are heated to their boiling point
(∼ 4300 K) (Gao et al., 2007). The peak value of the incandescent
signal was converted to the rBC mass based on a calibration curve determined
using fullerene soot (FS) particles (stock 40 971, lot: L20W054, Alfa Aesar,
USA). The calibration procedure used in quantifying rBC masses was the same
as that used in previous studies (Moteki and Kondo, 2010; Miyakawa et al.,
2016). The effective density function of FS was determined on the basis of a
DMA–aerosol particle mass (DMA–APM) system in Yutaka Kondo's laboratory at
the University of Tokyo and was consistent with previous results (Moteki and
Kondo, 2007; Gysel et al., 2011) (Fig. S1 in the Supplement). In our study,
the mass of the individual FS particles ranged from 0.35 to 89.5 fg, which
corresponds to 80–700 nm in mass equivalent diameter (MED), assuming an rBC
density of 1.8 g cm-3 and ideally spherical particles. Due to
variations in morphology and composition under ambient conditions, the
incandescent signal may not always be linearly proportional to the mass of
rBC particles. The uncertainty of the derived rBC mass on the basis of the
incandescent signal was estimated to be ∼ 30 %, and the uncertainty
of the derived MED values was estimated to be ∼ 10 %. The mass size
distribution of the rBC in the OBB plume typically peaked at 180–200 nm
(< 10 fg). Extrapolation using a lognormal function fit to the
observed size distribution suggested that the missing rBC particles with
MED < 80 nm and MED > 500 nm only cause minor mass
underestimation (∼ 5 %). The size-dependent detection efficiency of
SP2 for FS particles was also evaluated on the basis of a DMA-SP2-CPC
(model 3010, TSI Inc., USA) system, and we found that the SP2 detection
efficiency was in the range of 0.94–0.98 for particles with MED values
larger than 80 nm (shown in Fig. S2).
The scattering signal of the SP2 was calibrated using PSL particles with
known sizes (170, 200, 254, 300, 500 and 1000 nm). It is worth noting that
only PSL particles with sizes larger than 166 nm can be detected adequately,
implying that the ambient measurement of the SP2 will underestimate the total
number concentration of light-scattering particles because the particles that
are smaller than this size threshold are not counted. A two-elemental
avalanche photodetector was employed in the SP2 to determine the actual
position of the particle in the laser beam, which allows for delay time and
coating thickness analysis of rBC particles with a core–shell structure (Gao
et al., 2007). Detailed information on the SP2 is provided in the literature
(Schwarz et al., 2008, 2010; Moteki et al., 2014).
Determination of the coating thicknesses of rBC-containing
particles
As mentioned above, when the rBC-containing particles pass through the laser
beam, the rBC component needs a short period of time to absorb energy to
gradually evaporate the coating materials. This means that the time when the
rBC reached its boiling point was later than the time of occurrence of the
peak of scattering. For the rBC-containing particles with core–shell
structures, the coating thickness can be semi-qualitatively represented by
the delay time of the LII peak (Δt), which is defined as the elapsed
time between the occurrence of the peak of the scattering signal and the peak
of the incandescence signal; a positive value of this quantity indicates that
the peak of the incandescence signal occurs after the peak of the scattering
signal. In principle, the larger the Δt value is, the more likely the
rBC core is to be thickly coated. Such phenomena have been frequently
reported in a number of studies (Moteki et al., 2007; Subramanian et al.,
2010). In our experiment, a histogram analysis of the Δt value
demonstrated that there was a small Δt∗ peak at 0.8 ± 0.5
(2σ) µs for the uncoated FS particle with
MED < 400 nm. Apparently, this Δt∗ peak did not
result from the coating effect, but the intrinsic error of the photodetector
of the SP2. We regarded the particles with Δt values larger than 1.3
(mean + 2σ) µs as being coated. A detailed classification
of thinly coated and thickly coated particles is discussed in Sect. 3.4. In
practice, estimation of the coating thickness of rBC cores in terms of
Δt values is sometimes problematic because, first, the Δt
values showed a discontinuous increase with an increase in coating thickness,
depending on both the rBC core and the coating material. A laboratory study
of graphite particles coated with organic liquids indicated that the Δt value jumped from less than 1 to 3 µs after the coating
thickness of particles exceeded a threshold value (Moteki and Kondo, 2007).
Such variations provide only a measure of the minimum detectable coating
thickness, and they do not permit precise estimation. Second, the Δt
value can be negative in cases where the peak of the incandescence signal
occurs before the peak of the scattering signal (Sedlacek et al.,
2012). Negative Δt values have been reported not only in biomass
burning plumes (Sedlacek et al., 2015) but also in laboratory experiments
(Moteki and Kondo, 2007). An accepted explanation is that the rBC component
is located at or near the surface of non-refractory matter (such particles
are said to belong to the non-core–shell type or the attached type). When
such particles passed through the laser beam, the rBC did not absorb a
sufficient amount of energy to evaporate the non-rBC substances, and the
occurrence of fragmentation allowed some of the remaining non-rBC substances
to pass though the laser, producing a scattering signal after the
incandescence signal was induced. Moteki et al. (2014) suggested the use of
the time-dependent variation of the scattering cross section (Cs)
to differentiate the coated rBC particles from the attached type and proposed
a measurable parameter, a logarithm of the ratio of Cs before
evaporation (Cs-be) to Cs at the onset of incandescence
(Cs-oi), log(Cs-be/Cs-oi), to quantify the
contributions from the different types.
Following the same principle, we calculated the coating thicknesses of
rBC-containing particles with the core–shell structure on the basis of the
leading-edge-only (LEO) fitting method (Gao et al., 2007). The physical
interpretation of this method has been described in the literature (Gao et
al., 2007; Laborde et al., 2012). In the LEO fitting approach, the laser
intensity profile of the SP2 was predetermined from an analysis of the PSL
particles, and the “leading edge” data are selected according to the
criterion of t < -2.5σ. Here, σ denotes the
standard deviation of the Gaussian function of the laser intensity profile.
Using a strict threshold (e.g., t < -3σ) can reduce the
risk of the onset of evaporation of the coating; however, it significantly
increases the possibility of incorrect Gaussian fitting. The optical diameter
of the undisturbed rBC-containing particles was estimated using Mie
scattering theory and the LEO-fitted scattering peak height, presuming a
refractive index of m = 1.5 - 0i for the coating materials. To
test the validity of the LEO fitting method, laboratory experiments were
performed using FS particles (Alfa Aesar 40971, lot: L20W054) coated with
oleic acid (molecular weight 282.46, boiling point 360 ∘C). The
instruments used in this procedure are the same as those described in the
literature (Moteki and Kondo, 2007). FS particles with mobility diameters of
100, 150, and 200 nm were selected using the DMA and then coated with oleic
acid using a heated oleic oil bath. The coated particles with shell mobility
diameters of 180, 200, 250, 300, and 350 nm were selected by a second DMA
and then measured using the SP2. We excluded the doubly charged particles
from the data analysis because both the core and shell sizes of these
particles were apparently larger than those of the singly charged ones. The
shell diameters of the coated particles were well determined using the LEO
fitting method. Linear regression analysis of the calculated and measured
shell diameters demonstrated a good positive correlation with a high
correlation coefficient (r2=0.9), as shown in Fig. S3. The coating
thickness of rBC was calculated using (Dp-Dc)/2,
where Dp and Dc are the shell and core diameters,
respectively, of the rBC-containing particles. The shell / core (S/C)
ratios were calculated using Dp/Dc. The total
uncertainty of the S/C ratio values was calculated to be 14 %.
Non-methane volatile organic compounds (NMVOCs)
In the present study, the mixing ratios of the NMVOCs in the gas phase in
the OBB smoke were measured simultaneously using a high-sensitivity
proton-transfer-reaction mass spectrometer (PTR-QMS 500, IONICON Analytik
GmbH) with a time resolution of 1.9 s. Four primary ions
(H318O+, NO+, O2+ and H+• (H2O)2) were used, and more than
20 product ions (NMVOC • H+) were selectively monitored, according to predetermined multiple ion
detection (MID) settings. The detection limit and dwell time of PTR-QMS for
NMVOCs were generally less than 0.3 ppbv and 0.1 s, which are fully
satisfactory for the measurement of OBB smoke. Detailed information on the
configuration of the instrument is provided in the literature (Inomata
et al., 2015).
Temporal variations in the mixing ratios of CO and CO2, and the
number concentrations of rBC and non-rBC particles for the burning of wheat
straw (a) and rapeseed plants (b). The yellow (dry
distillation step of the biomass), red (flaming-dominant combustion) and blue
(smoldering-dominant combustion) shaded areas in the plot represent the
different burning states.
Determination of combustion states and the rBC emission ratios
As mentioned, the excess mixing ratios of CO (ΔCO) and CO2
(ΔCO2) are normally used to estimate the modified combustion
efficiency (MCE) of biomass, which was defined as ΔCO2/(ΔCO2+ΔCO). In the present study, the mixing ratios of CO and
CO2 were measured at a time resolution of 1 s. Although the real-time
variation in MCE was obtained, comparisons were difficult because the
combustion states and durations varied significantly among the different
cases. Here, a fire-integrated modified combustion efficiency was calculated
on the basis of Eq. (1).
MCE=∑ΔCO2∑ΔCO+∑ΔCO2=∑CO2plume-CO2baseline∑COplumeCObaseline+∑CO2plume-CO2baseline,
where ∑ represents the time-integrated total mixing ratios of ΔCO and ΔCO2, and [X] is the mixing ratio of species X,
expressed as ppmv. Correspondingly, the emission characteristics of rBC were
indicated by the rBC emission factor ΔrBC / ΔCO, which was
calculated on the basis of Eq. (2).
ΔrBC/ΔCO=∑rBCplumerBCbaseline∑COplumeCObaseline,
where [rBC] is the mass concentration of rBC in unit ng m-3. The
baseline of the mass concentration of rBC and the mixing ratios of CO and
CO2 were determined from the linear interpolation of the data before and
after each combustion experiment.
Results and discussion
Evolution of the combustion process
Figure 1 shows two sample time series of the number concentrations of rBC and
non-rBC particles and the mixing ratios of CO and CO2 in the smoke for
a wheat straw combustion case (Fig. 1a) and a rapeseed plant combustion case
(Fig. 1b). The evolution of the concentrations of particles and gases were
similar, although the carbon content and combustion duration differed
between the two biomass types. This result indicates that the combustion
process was overwhelmingly important in determining the variations in
emission characteristics. As described in the literature (Andreae and
Merlet, 2001; Reid et al., 2005), smoldering combustion normally occurs after
flaming-dominant combustion ceases, resulting in two isolated peaks for the
mixing ratios of CO and CO2. However, we did not observe such a clear
boundary between these two stages in our burning experiments. This result
occurred primarily because the combustion durations were short (less than
200 s), and both flaming and smoldering combustion might occur
simultaneously in different parts of the biomass. The mixing ratios of CO
displayed a broader tail than those of CO2, implying that the relative
importance of smoldering combustion increased at the end; this effect is
particularly clear in Fig. 1b.
The temporal variations in the number concentration of rBC were highly
correlated with those of CO2. This phenomenon is in accordance with
previous conclusions that the production of rBC particles is mostly related
to flaming-dominant combustion processes (Pan et al., 2012). In the
high-temperature and low-oxygen environment present in the inner part of
flames, the evaporated unsaturated alkanes tend to pyrolyze to soot precursor
particles (i.e., PAHs), which is followed by the aggregation and considerable
growth of the rBC particles. The non-rBC particles appear to be mostly
emitted under low-temperature burning conditions (Reid et al., 2005).
Directly after the biomass was ignited, a small peak (yellow shading in
Fig. 1) in the number concentration was sometimes observed, which could be
related to the emissions from combustion of butane fuel of lighters. As
flaming temperature increased up to 800 K, rBC particles were produced in
significant quantities, and these particles provided substantial numbers of
condensation nuclei for semi-volatile compounds. As a result, the number
concentration of non-rBC particles decreased to almost zero (red shading in
Fig. 1). When combustion shifted from the flaming-dominant to
smoldering-dominant stage during the second half of the burning period (blue
shading in Fig. 1), the number concentration of non-rBC particles increased
again. This observation can be explained by the secondary condensation growth
of pyrolyzed compounds that occurred because the temperature was not high
enough to cause their complete oxidation. The number size distribution and
volume size distribution of non-rBC particles during burning experiments are
shown in Fig. S4. Overall, these observations confirmed the results of
previous studies that indicate that particle formation in OBB is essentially
a nuclei-limited condensation process, and rBC and CO2 were mostly
produced by flaming combustion, whereas organic matter and CO were emitted
from the smoldering combustion (Reid et al., 2005).
Description of sample types; overall combustion states; CO,
CO2, and rBC concentrations; MMD and geometric standard deviation
(gσ) of rBC; and emission ratios for the burning experiments. The
ordering of these quantities is the same as in previous studies (Inomata et
al., 2015).
No.
Samplea
MCE
Duration
COb
CO2
rBCc
rBC / CO
rBC / CO2
MMDd
gσ
type
[10th, 90th]
Second
ppbv
ppmv
ng m-3
ng m-3 ppbv-1
ng m-3 ppmv-1
nm
1
A
0.964 [0.941, 0.991]
164
1181
695
13 983
24.2
904.8
215
1.32
2
A
0.930 [0.909, 0.982]
122
311
91
2366
15.6
1171.5
188
1.46
3
A
0.952 [0.884, 0.973]
94
261
114
1446
11.3
570.4
152
1.44
4
A
0.949 [0.913, 0.999]
150
343
141
4812
28.7
1541.0
187
1.44
7
A
0.953 [0.830, 0.987]
123
256
114
1408
11.2
554.6
160
1.42
8
A
0.976 [0.960, 0.994]
184
380
340
6290
33.9
832.6
191
1.37
9
A
0.917 [0.900, 0.987]
121
189
46
168
1.8
165.3
187
1.47
10
A
0.944 [0.911, 0.979]
120
274
101
787
5.9
348.9
148
1.43
11
A
0.862 [0.828, 0.920]
125
464
64
470
2.1
331.8
152
1.44
12
A
0.937 [0.853, 0.988]
151
230
75
429
3.8
256.5
148
1.42
13
A
0.950 [0.896, 0.976]
143
282
118
1790
13.0
682.8
163
1.46
14
A
0.952 [0.837, 0.964]
158
290
126
1295
9.1
461.1
160
1.50
15
B
0.909 [0.881, 0.999]
220
4757
1045
11 255
4.8
484.5
196
1.33
16
B
0.904 [0.857, 0.999]
207
1339
277
8658
13.2
–
177
1.36
17
B
0.961 [0.840, 0.988]
97
75
41
250
6.8
276.2
148
1.45
18
B
0.884 [0.730, 0.999]
230
1373
230
5350
8.0
1045.7
181
1.41
19
C
0.943 [0.902, 0.999]
244
2702
983
19 802
15.0
906.0
204
1.29
20
C
0.923 [0.891, 0.999]
226
3512
926
13 402
7.8
651.2
189
1.33
21
C
0.909 [0.839, 0.947]
258
286
63
209
1.5
149.6
137
1.39
22
C
0.951 [0.895, 0.976]
188
956
408
3052
6.5
336.4
155
1.41
23
C
0.960 [0.874, 0.985]
172
457
241
1259
5.6
234.9
142
1.44
24
C
0.954 [0.944, 0.994]
191
846
386
4609
11.1
537.4
144
1.47
a Sample type: wheat straw/dry (A), wheat straw/wet
(B), rapeseed plant/dry(C); b the mixing ratio of CO was diluted
by 22 times; c the mass concentration of rBC was diluted by
46 times; d mode diameter in mass size distribution of rBC,
abbreviated as MMD.
Table 2 summarizes the information on the sample types, the mixing ratios of
CO and CO2, the mass concentration of rBC particles and MCE for all
burning experiments. Although the combustion proceeded generally from flaming
to smoldering phase, both flaming and smoldering combustion sometimes
occurred at the same time at different positions of fuel. In addition, the
durations of flaming or smoldering were different case by case. Here, the
fire-integrated MCE value was used to represent the overall combustion
condition for each combustion case. As shown, the averaged MCEs have no
significant differences for the combustion of dry wheat straw (0.86–0.98),
wet wheat straw (0.88–0.96), and dry rapeseed plants (0.91–0.96). The
average mass concentration of rBC after 46 times dilution ranged from 0.25 to
19.8 µg m-3, and the average mixing ratio of CO after
22 times dilution ranged from 95 to 5003 ppbv.
Variations in rBC size as a function of MCE
The masses of rBC particles as a function of mass equivalent diameter (MED)
displayed a lognormal distribution for all burning cases. For better
expression, each dM/dlogDp distribution was
normalized so that its maximum value equalled 1 (Fig. 2a). We found that the
mass mode diameter (MMD) for each combustion case clearly increased from
152 nm to 215 nm as the overall MCE value increased from 0.862 to 0.964.
The correlation (r2=0.59) was significant at the 95 % confidence
interval (Fig. 2b). This result indicates that flaming combustion tends to
produce larger rBC particles than smoldering combustion. It is consistent
with previous studies, which reported that rBC particles formed considerably
in intense flaming combustion due to less efficient transport of oxygen into
the interior flame zone. As a result, growth in the size of rBC particles was
rapid because the coagulation rate of particles is roughly proportional to
the square of their number concentration (Lee and Chen, 1984). In this study,
these consecutive processes were interrelated and could not be decoupled in
the analysis. Nevertheless, this mechanism explained the outliers (contained
within the dashed ellipse shown in Fig. 2b), in that the number concentration
of rBC particles was not sufficient to support substantial growth. For
smoldering combustion, the production of rBC precursors (i.e., PAHs) was less
effective because of the low temperature.
Normalized mass size distribution (maximum value = 1) of rBC
particles for each burning experiment (a). For better understanding,
lognormal curve fittings are shown for a flaming-dominant combustion case
(red circle) and a smoldering combustion-dominant case (blue circle).
Variation of mode mass equivalent diameter (MMD) as a function of the
modified combustion efficiency (b). The size of the circle indicates
the average rBC mass concentration for each burning case. Blue dashed line
and blue shaded area are the linear fitting and confidence interval for fit
coefficients. The data in the gray dashed circle are excluded in the linear
fitting because of their low rBC mass concentrations.
As mentioned, the MMD of rBC particles in OBB plumes was determined by the
combustion condition; however, its variations during the evolution of the OBB
combustion process have not been fully investigated, because the separation
of the combustion stage for in situ measurement was difficult. A recent
study (Taylor et al., 2014) reported that the MMD of rBC particles was
152 nm. This value is apparently smaller than the frequently presented
values (180–200 nm), and the authors attributed the differences to
nucleation scavenging processes during transport. Published studies on OBB
plumes are mostly based on airborne SP2 measurements. Andreae and
Merlet (2001) noted that airborne measurements tend to be biased toward
flaming combustion because the plumes formed during the flaming stage were
more likely to be injected to higher altitudes than those formed during the
smoldering stage (Kondo et al., 2011). Another explanation involved rapid
coagulation processes, in which small rBC monomers might easily form
agglomerates or clusters driven by organic coatings when the temperature of
the plume decreased. This process likely occurs so quickly (on the timescale
of seconds) that ambient measurements (OBB plume age > 1 h)
cannot detect this process.
The variations in the emission ratios of rBC and ΔrBC / ΔCO, as a function of averaged MCE for all burning cases.
Previous observations and the results of laboratory burning experiments are
displayed in the plot.
Emission ratio of rBC particles
Figure 3 shows the dependence of the ΔrBC / ΔCO ratio on
the MCE for all combustion cases. For comparison, the results from previous
OBB experiments in the laboratory, as well as from field measurements and
emission inventories, are also plotted in the same figure. As shown, the
ΔrBC / ΔCO ratio increased from
1.5 ng m-3 ppbv-1 to 34 ng m-3 ppb-1 as the MCE
increased from 0.91 to 0.98. The results of fitting a power function were
similar to those from previous studies (McMeeking et al., 2009; May et al.,
2014), even though different types of biomass were combusted. This result
indicates that the MCE value is a key parameter for determining the rBC
emission intensity from OBB, irrespective of the difference in the types of
biomass used. In the present study, we tested two biomass conditions (dry and
wet) for wheat straw. For combustion cases involving wet wheat straw, we
found that the values of the ΔrBC / ΔCO ratio were always
less than 7.1 ng m-3 ppbv-1 as the MCE value increased up to
0.96, and these values are much smaller than that
(25.3 ng m-3 ppbv-1) corresponding to dry wheat straw. This
result implies that the wet biomass was unfavorable for the production of rBC
particles. This phenomenon is consistent with the experimental results
described by Chen et al. (2010), who reported an evident decrease in the
emission factor of elemental carbon and a moderate increase in the emission
factor of CO for burning of moist wildland biomass.
In this study, the average ΔrBC / ΔCO ratio was
13.9 ± 10.1 ng m-3 ppbv-1 for the burning cases with a
fire-integrated MCE > 0.95. This value was probably a low
estimation since both flaming and smoldering combustions were included in the
calculation. However, by selecting the cases with both the 10th and 90th
percentiles' MCE value > 0.90, we found that the average ΔrBC / ΔCO ratio was
23.1 ± 11.4 ng m-3 ppbv-1, higher than the value reported
from airborne measurements (8.5 ± 5.4 ng m-3 ppbv-1) for
the outflowing aged OBB plumes observed in Asia during ARCTAS-A (Kondo et
al., 2011) and the value (10 ± 5 ng m-3 ppbv-1) for
agricultural fires in Kazakhstan during ARCPAC (Warneke et al., 2009). In
these studies, the OBB plumes were sampled after they had undergone a week of
transport. In-cloud scavenging may take effect at these relatively low
values, although precipitation had not occurred. The ΔrBC / ΔCO ratios derived from the FLAME-I and II experiments
(McMeeking et al., 2009) are smaller
(8.3 ± 9.7 ng m-3 ppbv-1 as converted from emission
factors, expressed as g species kg-1 dry fuel), probably because the
combustion was inclined to smoldering (average MCE = 0.92). In situ
measurements of OBB plumes at the location where our samples for burning were
collected (Pan et al., 2012) indicated that the ΔEC / ΔCO
ratio from flaming-dominant burning was
17.4 ± 5.2 ng m-3 ppbv-1. This result approximates those
obtained in the laboratory. Measurements of OBB plumes in the North China
Plain found that the ΔEC / ΔCO ratios from wheat straw
burning ranged from 15 to 17 ng m-3 ppbv-1 (Pan et al., 2013).
Kondo et al. (2011) reported ΔrBC / ΔCO ratios as low as
2.86 ± 0.35 ng m-3 ppbv-1 (MCE = 0.96) for a fresh OBB
plume in North America. These large discrepancies might result from
significant rBC losses during transport. The yellow shading in Fig. 3
indicates the variability of ΔrBC / ΔCO ratios used in
emission inventories. Comprehensive analyses including all kinds of OBB
showed that the ΔrBC / ΔCO ratios were
8.6 ± 1.2 ng m-3 ppbv-1 (as converted from emission
factors, assuming a molar volume of 22.4 L at standard temperature and
pressure conditions for CO) (Andreae and Merlet, 2001),
7.5 ± 1.3 ng m-3 ppbv-1 (Akagi et al., 2011), and
9.0 ± 1.6 ng m-3 ppbv-1 using the bottom–up method (Yan
et al., 2006). Synoptically speaking, a relatively high rBC emission ratio
was suggested for estimating emission inventories of OBB because the majority
of rBC emissions normally occur during the flaming combustion stage, despite
the longer duration of the smoldering stage.
The emissions of rBC relative to ΔCO2 were also calculated for
each burning case (see Table 2). In general, the ΔrBC / ΔCO2 ratios did not show a clear increase with increasing MCE values,
and they varied from 149 to 1541 ng m-3 ppmv-1, with a mean value of 592.5 ± 364.1 ng m-3 ppmv-1. Schwarz et al. (2008)
reported a high ΔrBC / ΔCO2 ratio of 1770 ± 400 ng m-3 ppmv-1, much higher than the values obtained in this study. Small
values were also reported for OBB plumes observed in North America (100–357 ng m-3 ppmv-1), Siberia (167 ng m-3 ppmv-1)
(Kondo et al., 2011), and China
(245 ng m-3 ppmv-1) (Pan et al., 2012). Despite the differences
among these values, the ΔrBC / ΔCO2 ratio remains a
useful parameter in constraining the uncertainty of emissions of rBC from
OBB to within an order of magnitude in models.
Delay time of incandescence
For the rBC-containing particles, the delay time (Δt) of the peak of
the incandescent signal after that of the scattering signal is
widely accepted as a proxy for the coating thickness of rBC particles.
Particular caution should be employed when this concept is applied in data
exploration. First, SP2 cannot detect particles with shell diameters less
than 166 nm (the lower detection limit of SP2) owing to their weak
scattering signal. A systematic bias always occurs when investigating the
coating of rBC-containing particles with small rBC cores (i.e., MED less than
100 nm) because only very thickly coated rBC particles can be counted. To
avoid this bias, we report only the rBC-containing particles with relatively
large rBC cores (MED = 200 ± 10 nm). Second, only the
rBC-containing particles with positive Δt values were technically
deemed as having a core–shell structure, to which the delay time-based
method and LEO fitting method could be appropriately applied. In general, the
scattering profile of rBC particles in the core–shell structure contained a
main peak with a shoulder peak, which resulted from the coating material and
the rBC core, respectively. If there was only one predominant scattering peak
with a quasi-Gaussian shape, it suggested that the evaporation of non-rBC
coatings was insignificant, and the rBC-containing particle likely belonged
to the attached type, as demonstrated in the literature (Moteki et al.,
2014). In this study, the shell diameters of rBC-containing particles with
the core–shell structure were estimated using the LEO fitting method
(described in Sect. 2.2.2). We found that the shell diameters of
rBC-containing particles with rBC cores having MED = 200 nm ranged from
210 to 400 nm, with a 5th percentile value of 218 nm and a 95th percentile
value of 330 nm. The corresponding shell / core (S/C) ratios were
between 1.09 and 1.7.
The dependencies of Δt on the derived S/C ratios of all of the
coated rBC particles for all of the burning experiments are shown in Fig. 4.
The color in the plot represents the total number count of particles in each
bin. In general, Δt increases as the S/C ratio increases,
reflecting the fact that the rBC particles must spend a longer period of time
absorbing energy to evaporate thicker coatings. Histogram analysis showed
that both the S/C ratio and Δt displayed neither simple Gaussian
nor lognormal distributions. Instead, a multiple-peak Gaussian provided a
good fit to their number distributions (Fig. 4). For the distribution of the
S/C ratio, there were two modes. One mode (no. 1) occurs at an S/C
ratio = 1.18, and another mode (no. 2) occurs at an S/C
ratio = 1.34, indicating that the rBC particles had different levels of
coatings. The differences in coating thickness were most likely related to
the combustion state of biomass burning. As mentioned, although flaming
combustion emitted a substantial amount of semi-volatile organic carbons, the
production of rBC particles was also significant at high temperatures. The
competing condensing processes under nuclei-rich conditions resulted in
relatively thin coatings on the rBC particles, instead of direct formation of
particulate organic matter. Previous studies (Kudo et al., 2014; Inomata et
al., 2015) also found that the emission factor of NMVOCs during the flaming
stage was lower by an order of magnitude than that during the smoldering
stage. It also supported our conclusion that the thinly coated rBC particles
associated with mode no. 1 and mode no. 2 were primarily related to the
flaming combustion and smoldering combustion stages, respectively. In the
present study, the integrated area ratio between mode no. 1 and mode no. 2
was 0.76. This result suggested that the thinly and thickly coated rBC
particles were almost the same. It was worth noting that the histogram of the
delay times of rBC-containing particles also had two modes, with one mode
occurring at Δt = 1.74 µs and another peak occurring at
Δt = 3.18 µs. The integrated area ratio between these
two modes was 0.78, almost the same as the ratio derived from the S/C mode.
This result demonstrates that the rBC-containing particles with rBC cores of
200 nm and S/C ratios of 1.18 and 1.34 corresponded to delay times of 1.74
and 3.18 µs, respectively, at least in this study. Moteki et
al. (2007) investigated the relationship between delay time and coating
thickness for ambient rBC particles with MED = 200 nm, and they reported
that the delay time increased linearly from 0–1 to ∼ 4 µs as
the coating thickness increased up to 200 nm (S/C ratio = 2),
consistent with our study. It should be noted that the delay time for
uncoated rBC particles (S/C = 1) was not necessarily zero because of
intrinsic differences among SP2 instrumentations. For instance, a shift of
Δt∗ = 0–0.6 µs was found for uncoated rBC
particles among studies (Moteki et al., 2007; Sedlacek et al., 2012). By
subtracting the Δt∗ value (0.8 µs) in this study, the
Δt was found to be 0.9–2.4 µs for coated rBC particles in
the biomass burning plumes.
The dependence of the delay time of the peak of incandescence signal
after that of the scattering signal as a function of the
shell / core ratio for the rBC particles with MED = 200 ± 10 nm
for all the burning cases, and the multiple Gaussian fitting for all the data
of cross sections along the x-axis and y-axis. For comparison with the
curve fitting results, observational data of the flaming-dominant
(MCE > 0.95, red circle) and smoldering-dominant
(MCE < 0.9, blue circle) cases are also shown in the figure.
Coating thickness of rBC as a function of MCE
Figure 5a depicts the variations in modal Δt values for rBC particles
with MED = 200 nm as a function of the MCE value. As shown, Δt
clearly decreased as the MCE value increased (r2=0.63); the decreasing
trend was statistically significant at a level of 5 %. S/C ratios of
rBC particles (Fig. 5b) were found to be 1.4 at MCE < 0.9
(smoldering combustion) and 1.2 at MCE > 0.95 (flaming
combustion). Such a tendency was mostly due to formation of organic matter at
different combustion phases. Collier et al. (2016) reported that organic
aerosol emissions had negative correlations with MCE, implying that coating
processes of semi-volatile organics played a key role in the increase in the
S/C ratio. Variations in both the Δt values and the S/C ratios as
a function of MCE showed similar tendencies for particles with MED of the rBC
cores ranging between 190 and 250 nm. Airborne measurements during the
ARCTAS campaign (Kondo et al., 2011) showed an increasing tendency for the
values of the S/C ratio (1.3–1.66) with an increase in MCE, and the
authors explained that this phenomenon was because flaming phase plumes were
more aged than the smoldering plumes; a 205 ± 40 % increase in the
volume of the coating materials resulted in a larger S/C ratio than that of
rBC particles in smoldering plumes.
Brief summary of the coating thickness and shell / core (S/C)
ratios for rBC emissions from different sources collected from recent
studies.
rBC source
Coating thickness
S/C ratio
rBC core size
Age
Sampling description
Study
Biomass
Brush fires
65 ± 12 nm
–
a190–210 nm
0.5–1.5 h
Airborne SP2 measurements during
Schwarz et
burning
2006 Texas Air Quality Study
al. (2008)
∼ 15 nm
200 nm
–
Field measurements using SP2 in the agglomeration of Paris as part of the MEGAPOLI European project
Laborde et al. (2013)
Boreal forest
50–100 nm
2.0–2.5
152–196 nm
1–2 days
Airborne SP2 measurement during the second phase of the BORTAS project over eastern Canada and the North Atlantic during July–August 2011.
Taylor et al. (2014)
Agriculture
–
1.3–1.6
b120–140 nm
1–2 h
Airborne SP2 measurements during ARCTAS in spring and summer
Kondo et al. (2011)
Wheat, rapeseed
20 nm
1.2–1.4
200 ± 10 nm
< 10 s
Burning experiments in combustion
This study
plant
(11–54 nm)
chamber in laboratory environment
Asia continental
–
1.6
200 nm
2–3 days
Ground-based SP2 measurements at
Shiraiwa et
Free troposphere
–
1.3–1.4
200 nm
12 h
Fukue, Japan
al. (2008)
Aged air mass
44 nm
–
200 nm
Field measurement using SP2 in the agglomeration of Paris as part of
Laborde et al. (2013)
Traffic influence
2 ± 10 nm
200 nm
the MEGAPOLI European project
Traffic emission
110–300 nm
80–130
Highly aged
Ground-based SP2 measurement at an urban site in Shanghai, China
Gong et al. (2016)
Traffic emission
–
1.6–2.4
–
–
Clean Air for London (ClearfLo)
Liu et
Solid fuel burning
–
< 1.2
–
–
experimental campaign in winter, 2012
al. (2014)
Europe continental
–
1.45–1.6
–
–
Urban emission
20–30 nm
> 200 nm
1–2 days
Airborne SP2 measurement during the MILAGRO campaign
Subramanian et al. (2010)
a Volume equivalent diameter; b count
median diameter.
Statistically, the modal coating thickness of rBC particles was found to be
∼ 20 nm (Fig. S5). In fact, discrepancies exist among studies due to
differences in biomass types, burning conditions and sampling locations. For
example, ambient measurements in Europe indicated a coating thickness of rBC
particles of 15 nm on average, and more than half of the rBC particles had a
coating thinner than 10 nm (Laborde et al., 2013). Airborne measurements
reported a thicker coating (65 ± 12 nm) for rBC particles from brush
fires (Schwarz et al., 2008).
Variations in the delay time (a) and the shell / core
ratio (b) as a function of MCE values for rBC particles with
MED = 200 ± 10 nm.
Variations in the shell / core ratios of rBC particles with
MED = 200 ± 10 nm as a function of the emission factor of each
experiment. Here, EF is defined as the amount of each compound released per
unit amount of dry fuel consumed. The red, green and blue colors indicate the
dry wheat straw, wet wheat straw and rapeseed plant samples, respectively.
Table 3 summarizes recent studies that report the coating thicknesses and
S/C ratios of rBC particles. Among the studies, the coating thicknesses of
freshly emitted rBC particles from burning of wheat residues were smallest,
with a mean value of ∼ 20 nm. Coating thicknesses of rBC particles
from brush fire (Schwarz et al., 2008) and boreal forest (Taylor et al.,
2014) were 65 ± 12 and 50–100 nm, respectively. We noticed that there
was a large difference in the age of OBB plumes, and coating thicknesses of
rBC seemed to increase as they experienced a longer transport period. It
implied that aging of particles also played an important role in determining
the coating thickness of rBC particles. Further observational investigation
on the evolution of OBB process is also needed to explicate the contribution
to the total variability. The coating thickness of rBC particles from traffic
emissions varied significantly, depending on the urban emission and
photochemical processes. The transport pattern and meteorological parameters
(such as RH) also play a role in changing the morphology of rBC-containing
particles. Fan et al. (2016) reported that the hygroscopic shrinkage effect
under high RH conditions led aggregated soot particles to become more tightly
clustered, which resulted in an increase in the shell / core ratios. Such
phenomena have also been observed under ambient conditions (Adachi et al.,
2010). It has also been reported that condensation and coagulation processes
cause the voids of rBC aggregates to be filled or cause the particles to
collapse into a compact shell–core structure (Zhang et al., 2008).
Relationship between the S/C ratio and NMVOCs
Scatter plots of the S/C ratios of rBC particles with core
MED = 200 nm plotted against the emission factor (EF, in unit
g kg-1) of each condensable NMVOC are shown in Fig. 6. It is obvious
that the S/C ratio increases with the EF for the burning of both dry wheat
straw and dry rapeseed plants. As discussed in a previous study (Inomata et
al., 2015), the high EF values of the NMVOCs are related to smoldering
combustion, in which the production of rBC particles is less effective due to
the low-temperature conditions. The condensation of semi-volatile organics
was evident under rBC nucleus-limited conditions, resulting in higher S/C
ratios. We also found that the S/C ratio from the laboratory burning of
rapeseed plants was apparently higher than that of wheat straw under the same
NMVOC emissions conditions. This difference is likely attributable to the
physical formation processes of rBC particles under high-temperature
conditions for different combustion types. For instance, intensive fires tend
to produce non-spherical rBC particles with chain-like structures, which
display uneven coatings of semi-volatile organics on the rBC particles;
however, for compact, quasi-spherical rBC particles, it is much easier to
form evenly distributed, thicker coatings. For wheat residues, the S/C
ratios (∼ 1.4) of rBC for wet samples were apparently higher than those
(12–1.3) of dry samples at the same EF. This result indicates that the
microphysical properties of the rBC particles varied under different biomass
burning conditions. To answer these questions, further analysis of individual
particles using electro-microscopy is needed.