Introduction
Organic aerosol (OA) is one of the main components of
atmospheric particulate matter (PM) (Kanakidou et al., 2005; Zhang et
al., 2007). Identification of the sources of OA has proven to be a difficult
task due to their diversity and the continuous chemical evolution of the
corresponding organic compounds. The aerosol mass spectrometer (AMS, Aerodyne
Research) provides continuous information (OA mass spectra) that allows the
identification of some OA sources. Additionally, the OA elemental ratios
(O : C, H : C, N : C)
can be calculated providing useful information about the average chemical
state of the OA (Aiken et al., 2008). Positive matrix factorization (PMF)
(Paatero and Tapper, 1994; Lanz et al., 2007) is often used to deconvolute
the AMS data into a linear combination of factors. The resulting OA factors
are associated with primary OA (POA), such as the hydrocarbon-like organic
aerosol (HOA) or oxidized OA (OOA), which in many cases is related to
secondary OA (SOA) (Zhang et al., 2007). Factors linked to biomass burning
emissions (BBOA), cooking organic aerosol (COA), and marine emissions (MOA)
have also been identified. The OOA has been further separated into factors
based on their degree of oxidation and volatility (Zhang et al., 2007;
Kostenidou et al., 2009, 2015; Sun et al., 2011; Ge et al., 2012; Mohr et
al., 2012; Crippa et al., 2013a, b).
COA has been found to represent 10–35 % of the total OA measured in
urban locations (Allan et al., 2010; Sun et al., 2011, 2012; Ge et al., 2012;
Mohr et al., 2012; Crippa et al., 2013a, b; Lee et al., 2015). In Greece the
COA levels have been estimated in two major cities: Athens and Patras. During
the summer the COA-related source (named HOA-2) was 17 and 14 % of the
total OA in Athens and Patras, respectively (Kostenidou et al., 2015). For
the winter the corresponding contributions were 16 % for Athens and
12 % for Patras (Florou et al., 2017).
Emissions from meat cooking may produce large amounts of aerosol up to
40 gkg-1 (Hildemann et al., 1991). The types of meat cooked
(chicken, beef, etc.), other food ingredients, and the cooking method affect
both the aerosol emission rate and the composition of the corresponding
particles (Rogge et al., 1991; Mohr et al., 2009; He et al., 2010). For
example, Allan et al. (2010) suggested that the oil used during meat frying
may contribute more to the emitted PM than the meat itself in urban areas in
the United Kingdom.
Meat cooking particles contain palmitic acid, stearic acid, oleic acid,
nonanal, 2-octadecanal, 2-octadecanol, and cholesterol (Rogge et al., 1991).
Schauer et al. (2002) measured the emissions from cooking with seed oils,
showing that this process is a source of n-alkanoic and n-alkenoic acids.
Allan et al. (2010) reported AMS spectra for several oils used for cooking,
showing similar spectra (with enhanced fractions of signal at m/z 41 and
55) with some COA factors reported in the literature.
Most of the AMS spectra from ambient measurements related to COA are
characterized by peaks at m/z values 41, 43, 55, 57, 69, etc. and have an
O : C ratio ranging from 0.08 to 0.21 (Mohr et al., 2009,
2012; Allan et al., 2010; He et al., 2010; Sun et al., 2011, 2012; Ge et
al., 2012; Crippa et al., 2013a, b; Hayes et al., 2013). He at al. (2010)
reported coefficients of determination (R2) of 0.95–0.98 among the
spectra of OA emissions from different types of Chinese cooking despite the
differences in ingredients and cooking methods. Mohr et al. (2009) compared
the spectra of OA produced by grilling of hamburgers and chicken without
skin. R2 values greater than 0.9 were found between these AMS spectra.
Mobile aerosol measurements indicate that commercial and residential cooking
contribute to enhanced OA concentrations (Elser et al., 2016).
Despite the previous efforts, there are a number of remaining questions
regarding the characterization of the emissions related to cooking practices.
Separation of COA from the HOA and other primary components is still a
challenge for the PMF analysis (Mohr et al., 2009; Kostenidou et al., 2015).
Ots et al. (2016) attempted to constrain COA emissions in the UK using the
AMS-PMF results. Furthermore, the fate of these primary emissions in the
atmosphere is still unknown. The reactions with ozone (O3) and OH
radicals may significantly alter these aerosols. Hearn et al. (2005) studied
the reaction of oleic acid particles with ozone and concluded that relatively
fast heterogeneous reactions occur at the surface of the particles. Dall'Osto
et al. (2015) reported different COA factors for a rural site in the Po
Valley, Italy, with one being associated partially with primary organic
aerosol components such as HOA and partially with secondary components. In
contrast, the second COA factor did not correlate with primary tracers.
Kostenidou et al. (2015) reported an HOA-2 factor for the summer measurements
in Athens and Patras, Greece, that appeared to be associated with cooking but
was quite different from the COA factor identified in winter in the same
areas by Florou et al. (2017). The reasons for the differences of the COA
factor spectra even when the cooking practices are the same in the two
seasons were not clear. Due to the mild climate in Greece, there is no
significant change in what is cooked during the different seasons as opposed
for example to cities in much colder climates.
The aim of this work is to characterize the particulate emissions of pork
charbroiling, an activity that is thought to produce large amounts of
OA. Smog chamber experiments were conducted in order to characterize the
fresh and aged meat charbroiling emissions. The resulting spectra were
compared to COA factors derived from ambient measurements in Greece during
different periods of the year in an effort to explain the apparent
differences in COA spectra derived from the PMF analysis.
Summary of smog chamber experiments.
Chamber
Initial PM1
Aging
Initial
Initial
Final
Final
Average
O3
θ angle
exp.1
concentration
procedure
O : C
H : C
O : C
H : C
OH
formed
initial vs.
(µgm-3)
(moleccm-3)
(ppb)
final
1
130
UV illumination (4 h)
0.11
1.91
0.27
1.80
–2
35
22
2
400
UV illumination (8 h)
0.10
1.91
0.30
1.76
2.6×106
47
27
3
450
O3 addition (42 ppb)
0.10
1.95
0.21
1.90
6.5×104
–
16
4
335
UV illumination (7.5 h)
0.10
1.97
0.27
1.85
1.4×106
38
25
5
540
None
0.09
1.97
0.09
2.00
6.4×105
0
2.6
1 The total duration of experiments 1–4 was 7–8 h and that of
Experiment 5 was 4.5 h.
2 In Experiment 1 no d-butanol was added, and thus no OH radical concentration is
reported.
Experimental procedures
Chamber experiments
A set of five smog chamber experiments were conducted in the ICE-HT
environmental chamber facility. This facility is composed of a
temperature-controlled smog chamber room (3mW×4.5mL×2.5mH) incorporating over 300 UV light
lamps (Osram, L36W/73) capable of producing a JNO2 of
0.6 min-1 (when all lights are turned on). The reactor had a
volume of 10 m3. A commercial charbroiler was used for the meat
cooking. Natural wood coal was purchased from local distributors. A butane
burner was used for the ignition of the coal. The charbroiler was placed
outside the laboratory and adequate time was allowed for the coal ignition.
Pork was purchased from the local market. The meat was cut into 2×2×1 cm pieces which were placed on wood sticks (length:
20 cm). Approximately 100 g of meat were used for each
souvlaki. This type of cooking is widely used in Greece, both in restaurants
and homes. Ten to 15 souvlakia were cooked for approximately 20 min.
Using a metal bellows pump (Senior Aerospace, model MB 602) a fraction of the
emissions was introduced into the 10 m3 Teflon (PTFE) chamber that
had been pre-filled with clean air. Insulated 0.375 in. copper tubing (less
than 2.5 m in length) was used to transfer the cooking emissions into
the chamber. The copper tubing used for the sampling was insulated and was
therefore heated by the exhaust vapors. We have confirmed that the metal
bellows pump, as expected based on its design, does not generate particles or
volatile organic compounds (VOCs). The PM1 losses in this pump have been characterized previously
(Kostenidou et al., 2013) using two scanning mobility particle
sizer (SMPS) systems for both ammonium sulfate and
ambient particles. The losses were less than 10 % for particles larger
than 150 nm, increasing to 30 % for 100 nm particles. The flow
rate for the transfer line was approximately 170 Lmin-1. The
temperature and relative humidity were in the range of 20–25 ∘C and
15–35 %, respectively. Air was sampled 1 m above the charbroiler
for approximately 10 min in order to achieve a concentration inside
the chamber of the order of 100–500 µgm-3. The conditions
of each experiment are shown in Table 1. The NOx concentrations
were in the range from 1.5 to 8 ppb. The NO2-to-NO
ratio ranged from 2 to above 10. The NOx analyzer used is sensitive
to other NOy compounds and thus its measurements represent an upper
limit of NO2.
A high-resolution time-of-flight aerosol mass spectrometer
(HR-ToF-AMS, Aerodyne Research Inc.) measured the non-refractory PM1
aerosol. The vaporizer temperature was set at 600 ∘C and the voltage
difference between the filament and the ion chamber was 70 V. The
V mode of the AMS was used in these experiments. A SMPS (classifier model 3080, DMA model 3081, CPC model 3787, TSI)
measured the particulate number size distribution. The sheath flow rate was
5 Lmin-1 and the sample flow rate was 1 Lmin-1. The
size range of the SMPS under this configuration is from 10 to 500 nm.
The 10:1 ratio provides more accurate size distribution measurements as the
instrument has a sharper transfer function but it reduces the measurement
range to 10–300 nm. Given the modest size accuracy requirements in
this study (a few percent), we selected to cover a larger size range instead.
A multiple-angle absorption photometer (MAAP, Thermo Scientific Inc.) was
used for the measurement of the PM1 particulate black carbon (BC).
Quartz filters, placed after a PM2.5 cyclone, were used to collect
samples of the emitted COA from directly above the charbroiler and from
inside the chamber at the end of selected experiments. These samples were
used for the measurement of the organic (OC) and elemental carbon (EC) by
thermal – optical analysis (Sunset Laboratory Inc., EUSAAR 2 protocol) and
for the analysis of the water-soluble organic carbon (WSOC). For the WSOC
extraction a P parameter value equal to 0.1 cm3m-3 was used
according to Psichoudaki and Pandis (2013). The P parameter expresses the
ratio of the water used for the extraction of WSOC per volume of air sampled
on the filter to be analyzed. Samples were collected on Teflon (PTFE) filters
in one experiment in order to estimate the particulate mass emission factor
from the charbroiling of pork. Air was sampled from the charbroiler
(just above the pork) at a rate of 225 Lmin-1 through a
custom-built exhaust line (100 mm id). A portion of these emissions were
sampled through a 0.375 in. line after passing through a PM2.5 cyclone at
4 Lmin-1. Known portions (25 g) of pork were
individually cooked until well done and the emissions generated were sampled
from the exhaust line.
The VOCs were measured by a proton-transfer-reaction mass spectrometer (PTR-MS, Ionicon
Analytik). The drift tube was operated at 600 V at a constant
pressure of 2.2–2.3 mbar. The flow rate was 0.5 Lmin-1.
Further information about the PTR-MS operation can be found in Kaltsonoudis
et al. (2016). Blank measurements were conducted prior to the introduction of
meat cooking emissions to the chamber in each experiment. A series of gas
monitors was used for the measurement of the mixing ratios of the nitrogen
oxides (NOx), ozone (O3), carbon monoxide (CO), and carbon
dioxide (CO2) (Teledyne models: T201, 400E, 300E, and T360,
respectively).
Aging experiments were conducted in order to simulate the evolution of the
freshly produced COA as it reacts with typical oxidants (O3 and
OH) in the atmosphere. UV illumination was used (JNO2=0.59 min-1) and the chemical evolution of the particulate and gas
species was monitored. In some experiments, ozone was added and the
ozonolysis of cooking emissions in the dark was investigated (Table 1). No OH
precursor was used in any of the experiments.
Ambient measurements
Ambient aerosol was sampled at the ICE-HT institute (8 km NE from the center
of Patras) during February 2012 (Kostenidou et al., 2013). This period
included Fat Thursday (16 February) during which meat is charbroiled
everywhere in Patras. The instrumentation used for the ambient measurements
is described in Kostenidou et al. (2013). Briefly, an HR-ToF-AMS was deployed
for the characterization of the non-refractory PM1 aerosol composition,
a PTR-MS was used for the VOCs, an SMPS for the size distributions, a MAAP
for the BC, and a series of gas monitors were used for the NOx,
O3, CO, and CO2 concentrations. All instruments
sampled from approximately 4 m above ground. A PM2.5 cyclone was
used in front of the MAAP.
Data analysis
For the HR-AMS data analysis, SQUIRREL v1.56D and PIKA v1.15D with Igor Pro
6.34A (Wavemetrics) were used, applying the fragmentation table of Aiken et
al. (2008). The O : C and H : C ratios
were estimated using the improved method of Canagaratna et al. (2015).
High-resolution PMF analysis (Paatero and Tapper, 1994; Lanz et al., 2007)
was performed using the HR-AMS data from the chamber experiments and the
ambient measurements. The PMF evaluation tool PET (Ulbrich et al., 2009) was
used for both cases. The Multilinear Engine (ME-2) through Source Finder
software (SoFi) (Canonaco et al., 2013) was also used for the analysis of the
ambient measurements to investigate the robustness of the corresponding
results of the PMF. In all cases we used the m/z of 12–200 at high
resolution as inputs.
The OH radical concentrations were estimated using isotopically labeled
butanol (1-butanol-d9, Sigma). The change of the concentration of the PTR-MS
m/z 66 was used to calculate the OH concentrations based on the
second-order reaction of d9-butanol with the OH radicals. The corresponding
reaction constant used is 3.4×1012 cm3molecule-1s-1 (Barmet et al., 2012). The wall
losses corrections for the particles inside the chamber were calculated
according to Pathak et al. (2007) assuming a first-order loss rate for the
mass concentration of the total OA. The loss rate constant was established
during the characterization period of each experiment prior to the beginning
of chemical aging. The wall rate constants obtained for the experiments were
in the range of 0.14–0.28 h-1 and the corresponding linear fits
had very high correlation coefficients. Losses of particles to the walls do
remove part of the OA from the air in the chamber and make it “invisible”
for our measurements. However, the observed chemical changes were relatively
fast, taking place mostly within a couple of hours. The corresponding
timescales for losses were 4–6 h, so our conclusions are quite
robust. This can be clearly seen, for example, in the dark ozonolysis
experiment where fresh COA is decreasing following the O3 addition
significantly faster than it is lost to the walls before aging begun.
However, the fact that we could not observe the corresponding potential
changes to the COA particles deposited on the walls introduces some
uncertainty in the results. While one would expect similar changes in these
deposited particles if mass transfer of oxidants and condensable material was
rapid enough, we cannot confirm this. However, the effect of wall losses of
particles on the observed SOA / POA ratio is expected to be from modest to
small. This has also been addressed by the work of Hildebrandt et al. (2009),
who discussed the extremes of the potential fate of particles deposited on
smog chamber walls.
Source characterization experiments
Size distribution and chemical composition of the fresh COA
Table 2 summarizes the composition of the fresh cooking aerosol for the five
chamber experiments. The emitted aerosol is dominated by organic compounds
(above 99 %) in all experiments. BC was on average only 0.3 % of the
PM1. This is consistent with the OC and EC filter analysis of the
PM2.5 aerosol that was sampled directly from the charbroiler. In these
samples, the EC content for the fresh cooking emissions was less than
0.6 % of the total carbon. McDonald et al. (2003) reported that EC
emissions from charbroiling and grilling of chicken and beef were
0.3–2.7 % of the total mass using charbroilers fueled by natural gas. Li
et al. (2015) also reported low EC emissions due to cooking in China
(1.8–10.7 % for meat roasting, 7.5 % for fish roasting, 6 % for
street snack broiling, 1.9 % for cafeteria frying, and 10.7 % for
cafeteria broiling). In that study the WSOC-to-OC ratio was 0.05–0.15,
indicating that the freshly emitted aerosol was mostly hydrophobic.
Composition (% mass) of the freshly emitted COA for the
laboratory experiments.
Experiment
Average
1
2
3
4
5
Organics
98.4
99.0
99.4
99.4
99.6
99.2±0.5
Sulfate
0.1
0.5
0.1
0.1
0.1
0.1±0.2
Ammonium
0.0
0.0
0.0
0.0
0.0
0.0±0.0
Chloride
0.5
0.0
0.1
0.1
0.1
0.2±0.2
Nitrate
0.2
0.1
0.2
0.3
0.2
0.2±0.1
BC
0.8
0.4
0.2
0.2
0.1
0.3±0.3
(a) Fresh COA mass spectrum, (b) aged COA mass
spectrum for Experiment 2 (8 h of UV), and (c) the difference
between the fresh and aged COA.
Figure 1a depicts the average HR-AMS mass spectra for the fresh meat
charbroiling emissions. The initial spectra in all five experiments were
similar with each other having angles θ of 0–7∘ (R2
ranging from 0.983 to 0.999). The comparison of the AMS spectra based on
θ angles was favored for the analysis of the results in the present
paper. Briefly, a θ of 0–5∘ shows an excellent match
between the two spectra (with an R2 ranging approximately from 1 to 0.99), a
θ of 6–10∘ shows a good match (with an R2 ranging
approximately from 0.98 to 0.96), a θ of 11–15∘ shows that
the two spectra have many similarities but they are not quite the same (with
an R2 ranging approximately from 0.95 to 0.92), and finally a θ from
16 to 30∘ indicates spectra from different sources though there is
some limited similarity (R2 ranging approximately from 0.91 to 0.73).
Values of θ higher than 30∘ suggest clearly different spectra.
The advantage of the θ angle use for mass spectra comparisons is that it
can represent better small differences than the R2 coefficient. For
example, small differences of 1–5∘ all correspond to R2=0.99.
The initial O : C and H : C ratios based
on Canagaratna et al. (2015) and based on Aiken et al. (2008) in parenthesis
were on average 0.10±0.01 (0.08±0.01) and 1.94±0.03 (1.80±0.03), respectively. The main peaks of the corresponding AMS spectra were at
m/z values 27, 29, 39, 41, 43, 55, 57, 67, 69, 71, 79, 81, 83, 91, and 95.
The majority of these fragments correspond to homologous chains free of
oxygen. The gas-phase CO2 contribution to the CO2+ signal
was corrected by sampling through a HEPA filter during the experiments. The
CO2 levels were in the 395–435 ppm range and did not change
significantly during the course of each experiment.
(a) SMPS number and AMS mass distributions versus
Dp and Dva correspondingly for fresh COA and
(b) SMPS volume and AMS mass distributions versus Dva
and Dp.
The number-mode mobility diameter (Dp) of the fresh COA measured
by the SMPS was 86±20 nm, while the mass-mode vacuum aerodynamic
diameter (Dva) measured by the AMS was 224±30 nm.
Figure 2a shows the fresh COA number and mass distributions versus
Dp and Dva correspondingly for Experiment 1.
Figure 2b shows the mass and volume distributions versus Dva and
Dp correspondingly for the fresh COA in the same experiment. The
AMS and the SMPS estimated aerosol mass concentrations were quite different
during all chamber experiments. The SMPS mass concentrations were lower by
approximately a factor of 5 for a density of 1 gcm-3 compared to
the AMS total concentrations. Thus, an additional chamber experiment was
conducted, where AMS and SMPS concentrations were compared to gravimetric
measurements of the concentrations of COA samples collected on Teflon
filters. For the same period the AMS mass concentration (CE = 1) was
600 µgm-3, the SMPS (assuming density 1 gcm-3)
was 100 µgm-3, and the filter-based concentration was
500 µgm-3. This intercomparison shows that the SMPS mass
concentrations assuming spherical particles are problematic probably because
the fresh particles emitted from charbroiling are non-spherical. Katrib et
al. (2005) reported that during the ozonolysis of stearic acid needle-shaped
particles were identified by transmission electron microscopy. SEM pictures
of fresh COA particles in our experiments also suggested that the particles
were not spherical. However, particles evaporate in the SEM so the proof is
not conclusive. Given that CE values less than unity would further increase
the disagreement between the AMS and the SMPS estimates and that the density
of COA should be less than 2 gcm-3, we estimate that the
non-spherical shape of the particles introduces an error of the order of 2–4
in the volume concentration estimated by the SMPS measurements.
COA emission rates
Gravimetric analysis of the samples collected from above the charbroiler
yielded an aerosol emission factor of 4 gkg-1 of meat cooked.
Hildemann et al. (1991) studied the emissions from hamburger cooking of
regular and lean meet either by frying or charbroiling and reported emissions
between 1 and 40 gkg-1. McDonald et al. (2003) determined the
emission ratios of meat cooking (hamburger, steak, and chicken) due to
charbroiling or grilling and reported emission rates in the range
4–12 gkg-1 (McDonald et al., 2003). These rates vary by more
than 1 order of magnitude not only because different types of meat were
cooked but also due to the different cooking procedures (charbroiling,
grilling, frying, etc.), and cooking specifics (well-done, medium, slowly
cooked, medium cooking time, etc.). Generally charbroiling emits more particles than
frying and also the emissions increase with increasing fat content of the
meat cooked. There is also additional variability related to where the meat
is placed with respect to the very hot surfaces (e.g., charcoal). In the
present study we tried to duplicate the cooking conditions and practices used
in Greece.
Emissions of volatile organic compounds
Several VOCs were emitted during cooking though their concentrations compared
to the PM were low. In most cases less than 1 ppb of a specific VOC
was emitted per 100 µgm-3 of PM. The aromatic species
(benzene, toluene, xylenes) were emitted in similar amounts
(0.1 gkg-1). Table 3 presents the emission factors for some of
the measured VOCs based on a COA emission rate of 4 gkg-1 of
meat. To the best of our knowledge there is little information about VOC
emissions from cooking. For example, Schauer et al. (1999) reported the
emission factors from meat charbroiling over a natural gas-fired grill. So
our major objective was to add to this limited literature. Based on the
emissions measured the SOA formation potential of cooking would be limited
compared to the primary emissions. This is consistent with the limited
additional SOA that we have observed experimentally as discussed in the next
section. There was no detectable decrease of the concentrations of the VOCs
measured by the PTR-MS during the characterization periods.
Emission factors (gkg-1 of meat cooked) for several
VOCs.
VOC
PTR-MS
Emission rate
m/z
(gkg-1)
Acetonitrile
42
0.01±0.00
Acetone
59
0.03±0.01
Isoprene
69
0.05±0.01
MVK and MACR
71
0.03±0.01
MEK
73
0.01±0.01
Benzene
79
0.09±0.02
Toluene
93
0.09±0.03
Xylenes
107
0.10±0.04
Monoterpenes
137
0.04±0.02
Chemical aging of COA
Significant changes to the COA spectrum were observed during its oxidation.
The initial and final O : C and H : C
ratios for all the experiments are reported in Table 1. For Experiment 1, in
which the emissions were illuminated for 4 h the
O : C increased from 0.11 to 0.27. For Experiment 2, where
the UV lights were turned on for 8 h, the O : C
reached 0.30. For Experiment 3, in which dark ozonolysis took place, the
O : C ratio reached 0.21 2 h after the addition of
40 ppb of ozone. For experiment 4 the O : C ratio
reached 0.27 after 7.5 h of exposure to UV. No change was seen for
the O : C and H : C ratios for Experiment
5 in which the COA was left in the chamber without addition of oxidants or
exposure to UV.
For the experiments in which UV illumination was used the
O : C ratio increased from 0.1 to 0.2 in less than
2 h. For Experiment 3, in which 40 ppb of O3 were
added, the O : C ratio increased from 0.12 to 0.18 in less
than 1 h. Figure 3a shows the temporal evolution of the
O : C ratios during the five smog chamber experiments. In
Experiment 3 (dark ozonolysis) an increase prior to the addition of ozone was
seen due to small amounts of ozone (approximately 9 ppb) initially
present in the chamber. Figure 3b presents the corresponding
H : C ratio evolution during the five experiments. A
reduction in H : C by 10 % or so was observed in all
experiments.
(a) O : C ratios and
(b) H : C ratios for the COA smog chamber
experiments. Time zero corresponds to the beginning of the aging process.
Evolution of the θ angle with the initial AMS mass spectrum
during aging.
Scatter plot for the fractions of signal for Experiment 2 (UV for
8 h) and Experiment 3. (a) f44 to f43 for
Experiment 2, (b) f55 to f57 for Experiment 2,
(c) f44 to f43 for Experiment 3, and (d) f55
to f57 for Experiment 3.
(a) Ozone and estimated OH radical (m/z 66 corresponds to
d-butanol) concentrations in Experiment 2 (UV illumination) and
(b) ozone concentration during Experiment 3 (dark ozonolysis).
Mass spectra for the two resulting factors of the PMF analysis for
Experiment 1.
Time series of the resulting factors from the PMF analysis of the
chamber experiments without corrections for losses to walls. (a) PMF
factors for Experiment 3 (O3 addition) and (b) PMF factors
for Experiment 4 (UV illumination).
Time series of the four PMF factors found for the measurement period
including Fat Thursday (16 February 2012).
Mass spectra of the four PMF factors found for the measurement period
including Fat Thursday (16 February 2012).
Angles θ between COA factors and the (a) fresh meat
charbroiling emissions (average fresh spectrum) and (b) aged
(average of the UV exposure experiments) meat charbroiling emissions.
The θ angles between the fresh and aged COA AMS spectra are
summarized in Table 1. Differences in the AMS spectra between the fresh and
aged COA were present throughout the m/z range. The fractional contribution
of m/z 44, f44, increased during the UV aging and the dark ozonolysis.
Figure 1b shows the aged COA HR spectrum for Experiment 2 (after 8 h
of UV illumination). After aging with UV for 4 h the θ angle
was 22∘ in experiment 1. The addition of ozone resulted in a
15∘ shift in 4 h. The relatively fast change in the AMS
spectra is noteworthy (Fig. 4). After 1 h of UV illumination a
θ angle of 12∘ was observed. For the dark ozonolysis
experiment a 15∘ change was observed 2 h after the ozone
addition. These results indicate that the COA emitted from meat charbroiling
can change rapidly after it is emitted either during the day (when it is
sunny) or during the night when moderate levels of ozone are available.
Figure 5a shows the fraction of m/z 44 (f44) and m/z 43 (f43)
as they evolve over time during Experiment 2 (8 h of UV). Most of the
COA factors reported in the literature fall in the lower left part of the Ng
triangle (Ng et al., 2011). After the oxidation process the system position
tends to move up as f44 increases. A similar trend was observed for the
f55 to f57 plot (Fig. 5b) where both fractions decrease due to
chemical aging. A similar behavior was observed during ozonolysis (Fig. 5c
and d).
The driving forces for these chemical aging processes were reactions with
O3 and OH radicals. Significant O3 production was observed in
the UV illumination experiments with at least 40 ppb of O3
produced after a few hours of illumination. During the first hour,
15 ppb was formed and after 2 h the ozone concentration
reached 25 ppb. At the same time the OH radical concentration
increased up to 5×106 moleculescm-3. Figure 6a depicts
the O3 and OH concentrations for Experiment 2 in which UV
illumination was used. Similar results were obtained for the rest of the UV
illumination experiments. Figure 6b shows the O3 evolution during the
dark ozonolysis experiment. After the initial addition of 40 ppb of
ozone, approximately 5 ppb was consumed during the first
3 h.
Net OA production due to chemical aging was limited. In the five experiments
OA mass enhancements (after corrections for particle losses) were less than
10 % of the mass prior to the perturbation. This small change in mass
strongly suggests that a lot of the observed chemical changes were probably
due to heterogeneous reactions. Such reactions can explain the significant
changes in composition (e.g., O : C) and the small
additional OA formation in these experiments. Kroll et al. (2015) observed
similar changes in aerosol chemical composition during exposure to OH but at
exposures that were more than 1 order of magnitude higher than those in our
experiments. However, studies of the ozonolysis of COA components like oleic
acid (e.g., Morris et al., 2002) suggest that the corresponding reactions
have timescales of as little as minutes. These literature results suggest
that ozonolysis was probably the dominating chemical aging pathway in our
experiments.
While VOC concentrations remained stable when no UV light or oxidants were
used, the concentrations of formaldehyde, acetaldehyde, formic acid, acetone,
acetic acid, and methyl ethyl ketone all increased during the chemical aging.
Approximately 15 ppb / 100 µgm-3 of COA of
acetaldehyde and 8 ppb / 100 µgm-3 of COA of
formaldehyde were produced during the exposure to UV and O3. The
increase of the concentration of these relatively small compounds suggests
that fragmentation of the mostly larger organic molecules emitted during meat
charbroiling is taking place. It is not clear whether these molecules are
products of the organics in the particulate phase (that is products of the
heterogeneous reactions) or whether they were produced in the gas phase.
The water solubility of the COA also increased during its chemical aging. The
WSOC / OC ratio for the fresh emissions was measured in each experiment
and was always low with values in the 0.05–0.13 range. The WSOC / OC
ratio after chemical aging was measured in three experiments, two after UV
illumination (experiments 1 and 2) and one after dark ozonolysis
(Experiment 3). In all these three experiments the WSOC / OC ratio
increased:
to 0.7 for Experiment 1, 0.85 for Experiment 2, and 0.55 for Experiment 3.
This shows that the WSOC / OC ratio of the aged COA is significantly higher
than that of the fresh emissions and that the COA became a lot more
hygroscopic as it aged.
HR-PMF analysis was performed for each chamber experiment separately. More
information is provided in the Supplement (Sect. S1, Figs. S1–S18). For
experiments 1–4 two factors were identified: a fresh and an aged factor.
Figure 7 shows the mass spectra of the two factors for Experiment 1. The mass
spectra of the fresh COA factors had an O : C ratio of
0.09–0.11 and they were very similar each other (R2 > 0.992, θ< 7∘). They were also close to the average fresh mass spectrum from
all five experiments (R2 > 0.96, θ< 10∘). The aged
factors had an O : C ratio in the range of 0.20–0.26
depending on the degree of oxidation. The AMS spectra of the aged factors
after exposure to UV (experiments 1, 2, and 4) were similar to each other
(θ ranging from 2 to 6∘). The corresponding angles between the
dark ozonolysis experiment (Experiment 3) and the UV exposure ones were
higher ranging from 8 to 14∘ as the dark ozonolysis factor was less
oxidized. Figure 8 illustrates the time series of the two factors for
Experiment 3 (O3) and Experiment 4 (UV).
Ambient measurements
Ambient aerosol was sampled outside the ICE-HT institute (8 km NE from the
city center of Patras) during February 2012 for a period of 2 days. This
period included Fat Thursday (16 February) during which meat is charbroiled
everywhere in Patras. The ambient measurements have been corrected for the
CE, applying the algorithm of Kostenidou et al. (2007), comparing the AMS
mass distributions to the SMPS volume distributions. The CE for these
multi-component particles was 0.76±0.07. Applying HR-PMF analysis (using
PET) on the AMS spectra for these 2 days of measurements, four sources were
identified. PMF solutions up to five factors were examined and evaluated,
while the tested fpeak range was between -2 and 2. More
information for the selection of the factors is provided in Sect. S2 and
Figs. S19–S23. One factor was related to OOA, while the other three factors
were attributed to primary emissions: transportation (HOA), burning of olive
tree branches (otBB-OA), and meat cooking (COA). Given the small dataset, the
stability of the solution was further investigated using ME-2 analysis (SoFi)
and applying a constrained solution for the HOA using the HOA mass spectrum
of Kostenidou et al. (2013) with a=0.1 (Sect. S3, Figs. S24–S28). There
was no significant change in the factors in the two solutions. More details
are given in Sect. S3 and Figs. S23–S26. The cooking mass spectrum and time
series did not change significantly with an R2 > 0.99 between PMF and
ME-2 solutions.
Figure 9 shows the mass concentrations of the four factors. During midday on
Fat Thursday, the organic mass concentration was 23.2 µgm-3,
representing 81 % of the PM1. For the same period the
cooking-related factor represented 85 % of the organic aerosol
(17.5 µgm-3), while for the day before and the day after the
COA factor represented only 5 % of the total
OA. From various
studies that were conducted in Greek cities, COA appears to be 15–20 %
of the OA (Kostenidou et al., 2015; Florou et al., 2017). The mass spectrum
of the COA along with the rest of the factors obtained by the PMF analysis is
shown in Fig. 10. The m/z values contributing significantly to the COA
factor were 39, 41, 43, 44, 55, 57, 67, 69, 71, etc., which are
characteristic of cooking OA found in previous studies (Ge et al., 2012;
Crippa et al., 2013a, b).
Figure 11 summarizes the angle θ between the mass spectra of the
fresh and aged meat charbroiling OA and the PMF COA factors from ambient
measurements from other studies. Depending on atmospheric conditions (oxidant
levels) the COA AMS spectrum can be different. This can be seen by the
comparison of fresh and aged COA in these experiments against the summer and
winter COA factors in Greece. Other factors that appear to drive variability
can include the PMF analysis itself (e.g., mixing with other sources), the
type of food cooked, etc. The fresh laboratory COA spectrum has many
similarities with the COA factors obtained in Athens and Patras during the
winter (θ angles of 13∘ in both cases) especially considering
that these are independent PMF results from different campaigns. In contrast,
the aged COA spectrum is similar to the cooking-related factors (HOA-2)
identified during the summer in both cities (θ angles of 9∘
for both). The spectrum of the cooking OA of Fat Thursday in Patras (a sunny
period with moderate temperatures) was quite similar with the aged meat
charbroiling aerosol. This demonstrates that COA ages relatively fast under
ambient conditions when the necessary oxidants are available (i.e., sunny
summer days) and as a result PMF analysis can distinguish only one factor
(the aged COA). Even though this conversion is rapid the COA does not reach
high-oxidation states comparable to those of OOA. In addition, our results
suggest that before performing PMF analysis using default COA spectra as
external factors, one should account for their potential chemical aging.
The aged COA spectrum from our chamber experiments is similar to the COA
factor reported by Sun et al. (2011) for the city of New York during summer
(Fig. 11). This is also true for the COA-related factor (CIOA) reported by
Hayes et al. (2013) for the 2010 CalNex campaign in Pasadena, CA. However,
the COA factors reported for Fresno (Ge et al., 2012) and Paris (Crippa et
al., 2013b) were in better agreement with the fresh COA reported in this
work.