Observations of wildfire-influenced air masses at MBO
Figure 2 provides an overview of the meteorological conditions, trace gases
mixing ratios, and aerosol concentration and composition during the sampling
period (25 July–25 August 2013). The summit air was cool (average
temperature of 11.2 ± 4 ∘C) and dry (average RH of 46 ± 21 %), although there were periods (e.g., 16 and 23 August)
when MBO was in low clouds and measured RH reached 98 %. Wind was
generally strong (average = 5.7 ± 3.4 m s-1) with a dominant
flow from the west and southwest direction, which provides suitable
conditions for long-range transport of fire smoke from northern California
and southwest Oregon. Indeed, the bivariate polar plots of total
NR-PM1, submicrometer aerosol light scattering at 550 nm (σ550nm), and CO (Fig. 1b, d, and e) calculated using the OpenAir
software (Carslaw and Ropkins, 2012) all show the highest
values at a wind speed of ∼ 13 m s-1 from the southwest
direction, where the major complex fires were located (Fig. 1a).
Time series of (a) temperature (T) and relative humidity
(RH); (b) wind direction (WD) colored by wind speed (WS);
(c) mixing ratios of O3 and NOx; (d) mixing
ratios of CO, NOy, and PAN; (e, f) mass concentrations of
NR-PM1 species and σ550nm in STP (T=273 K, P=1013.25 hPa); (g) NR-PM1 composition; and
(h) f60 (= C2H4O2+ / OA). The solid
and broken lines in panel (h) indicate f60=0.3 % and
f60=0.5 %, respectively.
The average NR-PM1 concentration during the entire sampling period was
15.1 µg m-3, and 93 % of it was contributed by organics
(Fig. 1c). However, aerosol concentrations and composition changed
dynamically. Clean periods of low concentrations of aerosol
(NR-PM1 < 10 µg m-3) and gas phase pollutants
(e.g., CO, NOy, and PAN) were observed for the first week of sampling
(25–30 July) and during 18–21 August (Fig. 2d–f). During these periods,
ammonium sulfate contributed up to 90 % of the NR-PM1 mass
(Fig. 2g), and the OA spectra showed low abundances of
C2H4O2+ (m/z=60.021) and C3H5O2+
(m/z=73.029), which are ion fragments of anhydrous sugar (e.g.,
levoglucosan) and HR-AMS tracers for BB (Alfarra et al., 2007). f60,
which is defined as the fraction of the signal at m/z 60 (mostly
C2H4O2+) in OA spectrum, was generally below 0.3 %
(Fig. 2h), indicating minimal BB influence during “clean” periods (Cubison
et al., 2011).
In contrast, the other periods were characterized by higher f60 (up to
2 %), elevated NR-PM1 concentration (up to
∼ 210 µg m-3), and larger OA fraction (generally
> 90 % of NR-PM1; Fig. 2e–g). In addition,
σ550nm (up to ∼ 670 Mm-1), CO (up to
∼ 700 ppbv), NOy (up to ∼ 6.5 ppbv), and PAN (up to
∼ 2.2 ppbv) all increased dramatically during high-f60 periods
(Fig. 2d–e). In fact, the time series of all these parameters correlated
tightly, with Pearson's r2 in the range of 0.66–0.94 (Fig. S5). These
observations highlight the frequent and significant impacts of wildfire
emissions on air quality and atmospheric chemistry in the Pacific Northwest
region during this study. Note that potassium (K) is frequently used as a
tracer for BB aerosol, and the presence of K in aerosol particles was clearly
observed during high-loading periods. However, K concentration in aerosol was
overall very low and noisy throughout this study (Fig. S6), indicating low K
contents in wildfire emissions in the western US. Similarly, Maudlin et
al. (2015) observed no strong enhancement of K in wildfire smoke originated
from California and Oregon and concluded that it is not a reliable tracer for
BB in this region.
Impacts of wildfires on regional aerosol characteristics
Changes of aerosol concentration and composition due to
wildfires
Using f60 as an index for the influence of BB emissions on OA
composition, we divided the entire campaign into three regimes: (1) “No BB”
for periods with negligible BB influence and f60≤0.3 %; (2) “BB
Infl” for periods with detectable BB influences and moderately elevated
f60 values (0.3–0.5 %); and (3) “BB Plm” for periods with
f60 > 0.5 %, indicating intense and less processed BB
events. Note that periods with very low OA concentrations
(< 1 µg m-3), e.g., 18–21 August, were classified
as No BB regardless of the nominal f60 values. The average (±1σ) f60 values were 0.18 ± 0.10, 0.43 ± 0.05 and
0.77 ± 0.29 % for No BB, BB Infl, and BB Plm periods, respectively
(Fig. 3 and Table S1). Similarly, the average mixing ratios of CO, a gaseous
pollutant released from combustion, increased from 87.8 ± 17.9 ppbv
during No BB to 121.4 ± 24.8 ppbv during BB Infl and
178.3 ± 68.8 ppbv during BB Plm periods.
Figure 3 shows the comparisons of gas and particle phase properties among the
three regimes to illustrate the strong effects that wildfires have on gases
and aerosol composition in the Pacific Northwest region. For example, the
average NR-PM1 concentration was only 3.7
(±4.2) µg m-3 during No BB but increased by ∼ 4 and
∼ 7 times, respectively, during BB Infl
(13.4 ± 7.1 µg m-3) and BB Plm
(25.7 ± 19.9 µg m-3) periods. Aerosol measured at MBO
during BB Plm periods was predominantly organic (94.6 % of NR-PM1
mass; Fig. S7c). The fraction of OA in BB aerosols may be fuel dependent; for
instance, high values have been reported for ponderosa pine smoke emissions
(99 %) (Lewis et al., 2009), and somewhat lower values have been reported
for forest fires in southwestern Amazon (93 %) (Artaxo et al., 2013) and
North America boreal forests (87 %) (Kondo et al., 2011), and
agricultural fires in west Africa (85 %) (Capes et al., 2008). Even lower
values were observed in eastern Mediterranean wildfires (51.4 %)
(Bougiatioti et al., 2014) and Asian fires (60 %) (Kondo et al., 2011).
Since temperate evergreen vegetation was likely the dominant fuel during this
campaign, the high OA / PM1 ratio observed in this study appears
consistent with those of ponderosa pine.
Box plots that compare f60 values, CO mixing ratios,
NR-PM1 species concentrations, OA elemental ratios, carbon oxidation
states (OSc), σ550nm, and mixing ratios of trace
gases among three aerosol regimes (No BB, BB Infl, and BB Plm).
The whiskers indicate the 90th and 10th percentiles, the upper and lower
boundaries respectively indicate the 75th and 25th percentiles, the lines in the boxes
indicate the median values, and the markers indicate the mean values.
In addition to OA, concentrations of nitrate, ammonium, and chloride all
showed substantial increases that correlated with wildfire impacts
(Figs. 2–3, S8; Table S1). Nitrate, in particular, displayed large temporal
variations that correlated with wildfire plume influences, and its
concentration in the BB Plm regime was on average ∼ 11 times greater
than the No BB regime. Nitrate appeared to be bulk neutralized based on
comparing the total molar equivalent of inorganic anions (i.e., sulfate,
nitrate, and chloride) to that of ammonium (Zhang et al., 2005) during
wildfire-influenced periods (Fig. S9a), and the signal ratios of NO+ to
NO2+ observed in particles during these periods (2.15 ± 0.006)
were very similar to the ratio measured for pure NH4NO3 particles
(2.2; Fig. S9b), indicating that nitrate was mostly in the form of
NH4NO3. Note that for high-organic-loading
(> 50 µg m-3) periods excess ammonium relative to
sulfate, nitrate, and chloride was frequently observed. A possible reason is
the presence of significant amounts of organic anions in aerosol. Indeed,
CO2+ (m/z=44) and CHO2+ (m/z=45) – ion fragments for
carboxylic acids – were found to dominate the HRMS of aerosol during periods
of high OA loading (Fig. S7f). Another possible reason is overestimation of
ammonium concentration. Biomass burning can emit significant amounts of
nitrogen-containing organic compounds, including amines. These compounds can
produce NHx+ ions in the AMS, although they generally produce
significantly more CxHyN+ ions (Ge et al., 2014). Tight
correlations between CxHyN+ ions and biomass burning tracers
(e.g., CO, C2H4O2+, and C3H5O2+) were
observed, suggesting that amino compounds were likely emitted from wildfires
in the western US. However, the low abundance of CxHyN+
(∼ 0.3 % of total organic signal) indicates that organic nitrogen
compounds unlikely had a noticeable influence on ammonium quantification
during this study. Sulfate, on the other hand, displayed milder temporal
variation with poor correlation with BB tracers (Fig. 2d–f), indicating that
forest fires in this region are not a significant source of sulfate aerosol.
Collier et al. (2016) came to a similar conclusion through examination of
aerosol enhancement ratios in transported BB plumes.
Significant enhancements due to wildfires emissions were also observed for
PAN and NOy (Fig. 3). However, the mixing ratios of NOx (mostly as
NO2) were comparable among the three regimes. As a result, the
fractional contributions of PAN and particulate nitrate to total NOy
both increased due to wildfire influence (Fig. S10). Considering that MBO was
hours downwind of wildfire sources during this study, this observation is
consistent with the findings of Akagi et al. (2012) that NOx emitted
from BB is rapidly converted to PAN and particulate nitrate during plume
transport, which reflects high levels of acetaldehyde in fire plumes (Akagi
et al., 2011). The influence of wildfire emissions on O3 at MBO
appeared to be complex (Fig. 2c).
The average O3 mixing ratio in both BB Infl (49.1 ppbv) and BB Plm
(47.3 ppbv) regimes was higher than during the No BB (44.7 ppbv) periods
(Fig. 3), suggesting that O3 production was enhanced in BB emissions.
Similar observations have been made previously, which indicate that O3
tends to peak downwind of fire sources as a result of the interplay of fire
emissions (precursors and reactants) and chemical reactions (Jaffe and
Wigder, 2012; Wigder et al., 2013; Briggs et al., 2016).
Influence of wildfires on organic aerosol chemical properties
In order to demonstrate the influence of wildfires on bulk OA chemistry at
MBO, the average HRMS of OA for each of the three regimes is shown in
Fig. S7. OA was generally highly oxidized under all three regimes, and the
O / C of OA generally decreased as BB influence increased. In addition,
ions larger than 100 amu (fm/z>100) contributed a larger fraction of
the total organic signal during BB Plm periods (11 %) than during No BB
periods (5 %), consistent with BBOA containing a larger fraction of
high-molecular-weight compounds (Ge et al., 2012a; Lee et al., 2016). OA in
No BB air masses had an average O / C of 0.84 (O / CAA;
i.e., O / C calculated with Aiken-Ambient method is 0.63) and H / C of 1.48 (H / CAA=1.29), in agreement with previous HR-AMS measurements of free-tropospheric
OA at mountaintop sites (e.g., Sun et al., 2009; Rinaldi et al., 2015). The
average O / C for BB Infl and BB Plm periods was 0.77
(O / CAA=0.60) and 0.69 (O / CAA=0.53),
respectively, substantially higher than previously reported O / C for
fresh BB emissions. For example, laboratory experiments reported
O / CAA in the range of 0.15–0.60 for POA from BB, depending
on fuel type, burning condition, and burn mass (Heringa et al., 2011; Ortega
et al., 2013). The high O / C observed for BB-influenced OA at MBO
indicates that they were likely a combination of primary and secondary
components, with the secondary portion having a substantial contribution to
the bulk OA.
Changes in OA chemical composition due to wildfires is further investigated
using the f44 vs. f60 plot (Fig. 4). All OA data showed a
progression where lower f60 values were associated with higher f44,
consistent with aging of BBOA observed both in laboratory studies and from
airborne measurements (e.g., Cubison et al., 2011; Ortega et al., 2013;
Jolleys et al., 2015). f44 during No BB periods spanned the range of
0.13–0.25 (mean = 0.17), due to the dominance of highly oxidized OA. BB
Plm data fell within the region defined by the BBOA measured previously
(Cubison et al., 2011; Ortega et al., 2013) and overlapped particularly well
with fire plumes sampled above the North America continent during the 2008
NASA Arctic Research of the Composition of the Troposphere from Aircraft and
Satellites (ARCTAS) mission and aged BBOA from controlled chamber open
burning of biomass (Cubison et al., 2011). Ambient fire plumes tended to have
higher f44 and lower f60 values than the POA from burning of
various fuels in chamber studies (Ortega et al., 2013), mainly due to
atmospheric aging. However, the mixing of transported BB smoke with more
oxidized background aerosols likely also contributed to the changes in
f44 and f60 observed for ambient BBOA. Furthermore, combustion
conditions might also play a role in how plumes map to the
f44 ∼ f60 space, as it has been shown in both ambient and
chamber laboratory studies that flaming-dominated fires for certain fuel
types can lead to higher f44 and are associated with lower f60 than
more smouldering fires (Weimer et al., 2008; Jolleys et al., 2014; Collier et
al., 2016).
Scatterplot of f44 vs. f60. The grey markers correspond to
the measured OA during this study, and the SRCF case study data are colored
by cumulative solar radiation (∑SR). In addition, the five OA factors
identified in this study are shown as solid squares, and the BBOA factors
reported in the literature where multiple BBOA factors were derived are shown
with different markers. The dashed red lines denote f60=0.003 and the
boundaries set for BBOA (Ortega et al., 2013). The brown oval encompasses
ARCTAS fire plumes sampled above the North America continent (Cubison et al.,
2011).
(a–e) Time series of OA factors and corresponding tracer
compounds. Organic ions are in organic equivalent mass;
(f–g) diurnal variations of OA factors (the whiskers above and
below the boxes indicate the 90th and 10th percentiles, the upper and lower
boundaries respectively indicate the 75th and 25th percentiles, the lines in
the boxes indicate the median values, and the cross symbols indicate the mean
values) with the diurnal cycle of mean water vapor in panel (i);
(k–o) HRMS of OA factors colored by eight ion families at
m/z < 180, and (k′–o′) UMR MS at
m/z > 180 for each OA factor. The elemental ratios of each OA
factor are shown in the legends of panels (k–o), with those
obtained using the AA method in parenthesis.
(a–e) Bivariate polar plots that illustrate the variations
of the concentrations of each OA factor as a function of wind speed
(m s-1) and wind direction; (f) average OA composition during
the sampling period; and (g) volatility profiles of OA factors,
sulfate, and nitrate, with error bars showing the standard deviation of the
calculated slope, i.e., mass fraction remaining.
Aerosol source apportionment and contributions of primary and secondary BBOA
at MBO
To gain further insight into the influences of different sources and
processes on OA concentration and composition at MBO, we performed PMF
analysis on the HRMS of all NR-PM1 species acquired during this study.
PMF is commonly applied to the organic mass spectral matrix to determine
distinct OA factors (Zhang et al., 2011, and references therein), but
conducting PMF analysis on the combined spectra of organic and inorganic
aerosols allows for deriving additional information, e.g., the distributions
of inorganic signals among different factors and the nominal acidity of the
factors, which benefits the interpretation of the sources, chemical
characteristics, and evolution processes of OA (Sun et al., 2012). For this
study, a total of five OA factors were identified, including three different
BB-related aerosol types – i.e., BBOA-1 (O / C = 0.35), BBOA-2
(O / C = 0.60), and BBOA-3 (O / C = 1.06) – and two distinct
OOA factors, i.e., a less oxidized OOA associated with boundary layer (BL)
dynamics (BL-OOA, O / C = 0.69) and a more oxidized low-volatility
OOA representing free-troposphere aerosol (LV-OOA, O / C = 1.09).
Unlike the two OOAs, the three distinct BBOA factors all showed high
correlations with CO (r2=0.70–0.86; Table S2) and displayed sporadic,
high-amplitude events with large enhancements in concentrations during
wildfire-influenced periods (Fig. 5a–c). In addition, the polar plots of all
the BBOAs showed clear concentration hotspots in the southwest direction at
high wind speed (Fig. 6a–c), indicative of their associations with wildfire
plumes originating from SW Oregon and NW California (Fig. 1). Nevertheless,
the three BBOAs are distinctly different in terms of mass spectral profiles
(Figs. 5k–m and S11), oxidation degrees, and volatility (Fig. 6g), likely
due to different extents of aging and/or processing pathways. Similarly,
previous studies reported the identification of multiple BBOA factors
representative of different degree of atmospheric processing (e.g.,
Bougiatioti et al., 2014; Brito et al., 2014) and varying combustion
conditions (e.g., Young et al., 2015, 2016). BBOA-1 and BBOA-2 looked more
similar to the fresher BBOA factors, while BBOA-3 was more similar to the
aged BBOA factors derived in Bougiatioti et al. (2014) and Brito et
al. (2014) in terms of mass spectral features (Fig. 4).
Among the three BBOA factors, BBOA-1 had the lowest O / C (0.35) and the
highest H / C (1.76) and f60 (2.2 %) (Fig. 5k). In addition, the
mass spectrum of BBOA-1 showed prominent signals of C2H3+,
CHO+, C4H7+, C4H9+, and C9H7+,
markers for chemically reduced aerosols, and a high abundance of ions larger
than 100 amu (fm/z>100=25 %; Fig. 5k and k′). The UMR spectrum
of BBOA-1 at m/z > 180 exhibited a “picket fence”
fragmentation pattern where groups of peaks have 14 amu separation,
suggesting the occurrence of molecules with hydrocarbon moieties containing
different units of the CH2 group. The time series of BBOA-1 correlated
tightly with those of C2H4O2+ and C4H9+
(r2=0.94 and 0.95, respectively; Table S2), tracers for primary
emissions. Furthermore, BBOA-1 appeared to have a strong point source SW of
MBO and peaked in association with high wind speeds, suggesting that it could
be associated with plumes experiencing shorter transport times relative to
plumes from equidistant fire sources (Fig. 6a). Together, these observations
suggest that BBOA-1 was primarily associated with fresher and less processed
air masses from BB sources. In addition, BBOA-1 was found to be semivolatile
(Fig. 6g), which is consistent with previous findings that a majority
(50–80 %) of the POA in BB emissions is semivolatile (May et al., 2013).
The semivolatile behavior of BBOA-1 also explains the high degree of
correlation between BBOA-1 and nitrate (r2=0.60; Fig. 5a and
Table S2), a secondary species that is often found to correlate with
semivolatile OOA (SV-OOA) (Zhang et al., 2011). However, despite being a
secondary component, nitrate displayed tight correlations with primary smoke
markers, i.e., C2H4O2+ and C3H5O2+, at
MBO (Fig. S12). Therefore, it appears that fast processing near the fire
sources led to the rapid conversion of NOx to more oxidized compounds
such as PAN and nitrate. Based on these results, we infer that BBOA-1
represents fresher BB emissions and might be a surrogate for primary BBOA. On
average, BBOA-1 comprised 20 % of total OA mass during this study
(Fig. 6f), suggesting that fresh BB emissions exerted a significant impact on
regional air masses.
The more oxygenated BBOA-2 (O / C = 0.60; H / C = 1.72)
accounted for an average 17 % of the total OA mass (Fig. 6f). Its mass
spectrum displayed characteristics of aged BBOA with lower abundances of
C2H4O2+ (f60=1.1 %), CxHy+ ions
(31 %), and ions > 100 amu (fm/z>100=17 %) than
BBOA-1 (Figs. 4l, l′ and S9b). BBOA-2 also showed a somewhat less volatile
profile than BBOA-1, especially at TD temperature < 150 ∘C
(Fig. 6g). In addition, the temporal trend of BBOA-2 displayed tight
correlations with tracers for carboxylic acids, e.g., CHO2+ and
CO2+ (r2 of 0.91 and 0.79, respectively; Fig. 5b and Table S2),
but lower correlations with nitrate, C2H4O2+, and
C4H9+. These results suggest that BBOA-2 was more chemically
processed and likely contained secondary products. Indeed, the polar plot of
BBOA-2 (Fig. 6b) displayed a more dispersed pattern of sources than that of
BBOA-1, with hotspots located in various directions. Nevertheless, the
occurrence of a high-concentration band at 5–15 m s-1 in the SW
direction suggests important BBOA-2 sources from similar distances and
locations to BBOA-1. The dispersed source features are further evidence that
BBOA-2 is more secondary in nature than BBOA-1 and is likely more aged.
BBOA-3 contrasts strongly with BBOA-1 and BBOA-2 in chemical composition. The
HRMS of BBOA-3 had a very low C2H4O2+ signal (f60=4×10-8), a relatively high intensity of CO2+ (f44=0.215) and a high degree of oxidation (O / C = 1.07; Fig. 5m), all
of which greatly resemble those of LV-OOA (Fig. 5o). However, the mass
spectra at large m/z's indicated distinct chemical differences
between BBOA-3 and LV-OOA (Fig. 5m′ and o′), as there appeared to be a
higher abundance of high-molecular-weight species in BBOA-3. In addition, the
temporal variation patterns of BBOA-3 and LV-OOA were dramatically different
(r2=0.07), and BBOA-3 closely correlated with CO (r2=0.86;
Fig. 5c and Table S2), whereas LV-OOA did not (r2=0.008). As shown in
Fig. 6, the polar plot of BBOA-3 showed a high-concentration band from SW at
a wind speed of 5–15 m s-1, which overlaps with the hot spot shown in
the BBOA-1 polar plot (Fig. 6a). These results indicate that BBOA-3 was
associated with wildfires and likely formed both through rapid processing
near the wildfire source and during transport to MBO. However, given that
humic-like substances (HULIS) are a known component of BB emissions and that
these substances resemble BBOA-3 in terms of AMS mass spectrum, high degree
of oxygenation, and low volatility (Dinar et al., 2006; Adler et al., 2011),
it is possible that a fraction of BBOA-3 was HULIS as well.
Another important characteristic of BBOA-3 is that it appeared to be composed
of some very low volatility compounds. As shown in Fig. 6g, ∼ 60 %
of its mass remained in the aerosol phase at a temperature of
200 ∘C. This observation is consistent with previous studies which
have observed the presence of low-volatility and extremely low volatility
BBOA materials in aged wildfire plumes (Lee et al., 2016; Paciga et al.,
2016) and in SOA produced from major organic gases from BB (e.g., phenols)
(Yu et al., 2016). It is important to note that the highly oxidized BBOA-3 on
average accounted for 31 % of the total OA mass during this study, which
implies that a significant fraction of the highly aged BBOA may appear
indistinguishable from OOA from other sources due to mass spectral
similarities (e.g., low f60 and high f44) and hence would lead to
an underestimation of the influence of BB emissions on a regional scale.
(a) Map of the Pacific Northwest with the location of MBO
marked by a black triangle. Open diamonds represent MODIS satellite fire dots
detected during 13–17 August 2013 and are sized by fire radiative power
(FRP). Twelve-hour HYSPLIT back trajectories of air masses arriving at MBO
from 14 August at 20:00 to 16 August at 09:00 PDT are colored by time of
arrival at MBO. Markers indicate 1 h interval; (b) cumulative solar
radiation (∑SR) and average RH for each trajectory; (c) wind
direction (WD) colored by wind speed (WS) measured at MBO; mixing ratios of
(d) CO and NOx; mixing ratios of (e) nitrate, PAN, and
NOy; (f) five OA factors; (g) OA composition; and
(h) average carbon oxidation states and f60 of OA during the
SRCF case study period.
BL-OOA and LV-OOA accounted for the remaining 32 % of total OA mass
during this study. These two OOAs were not associated with BB, as indicated
by low f60 (Fig. 5n and o) and a lack of correlation with BB tracers
(Table S2). BL-OOA was relatively oxidized (O / C = 0.69; Fig. 5n)
and appeared significantly less volatile than nitrate but more volatile than
sulfate (Fig. 6g). BL-OOA showed a distinct diurnal cycle highly resembling
that of water vapor (Fig. 5i), which is a tracer for BL upslope flow during
the daytime at MBO (Weiss-Penzias et al., 2006). Photochemical production of
OA in the early afternoon may also contribute to the daytime increase of
BL-OOA. Furthermore, the time series of BL-OOA correlated with
CH3SO2+ (Fig. 5d and Table S2), a signature ion for
methanesulfonic acid (MSA) (Ge et al., 2012b). MSA is typically associated
with marine sources but has been found to have terrestrial sources as well
(Ge et al., 2012b; Young et al., 2016). All these results suggest the
influence of BL dynamics on BL-OOA. In comparison, the LV-OOA was highly
oxidized (O / C = 1.09) with a pronounced CO2+ peak in the
spectrum (Fig. 5o). In addition, Fig. 6g indicates that LV-OOA shared a
similar volatility profile to sulfate, showing no sign of evaporation until
the TD temperature reached nearly 130 ∘C, consistent with LV-OOA
previously determined in other ambient studies (Huffman et al., 2009; Paciga
et al., 2016). The diurnal pattern of LV-OOA appeared to be rather flat
(Fig. 5j), and its polar plot had the most dispersed feature among all
factors (Fig. 6e). All these observations suggest that this factor is
representative of free-tropospheric aerosol.
A case study of the aging of BBOA in wildfire plumes
Based on MODIS fire hotspot information, the SRCF was continuously burning
from 13 to 17 August (Fig. 7a). Three-day HYSPLIT back trajectories suggest
that air masses arriving at MBO from 14 August at 22:00 to 16 August at
09:00 PDT passed over the SRCF (Fig. 7a), consistent with the observations
of persistent SW wind at MBO during this period (Fig. 7c). MODIS also
detected a few hotspots from the Whiskey Complex fire
(∼ 43∘ N, 122.8∘ W) intermittently on August 15, but
the fire was much weaker than SRCF as indicated by the lower fire radiative
power (FRP, Fig. 7a). We therefore assume that the emissions arriving at MBO
during this time period were from a single source and therefore consistent in
transport distance and fuel type. Combining MODIS fire hotspots and back
trajectories, we estimated that the transport time of SRCF plumes ranged from
8 to 11 h before being sampled at MBO.
In order to examine how atmospheric aging affects BBOA chemistry, we
calculated cumulative solar radiation (∑SR) and average RH over the
total transport time (from source to MBO) for each trajectory and plotted
them versus air mass arrival time in Fig. 7b. ∑SR denotes the total
amount of solar radiation that the smoke plumes were exposed to during
transport and can be used as an indicator for the extent of photochemical
aging assuming the plumes were optically thin. RH in the air mass history
was relatively stable; however ∑SR clearly varied throughout the
measurement period such that some BB plumes experienced more solar radiation
than others and some were transported exclusively at night. Furthermore, the
burn conditions were modestly constant during this period with an average
modified combustion efficiency (MCE) value of 0.88 (±0.03) for the BB
plumes that met the criteria for MCE calculation (Collier et al., 2016). Furthermore, the MCE values
showed no differences between nighttime and daytime plumes and did not
correlate with ∑SR (Fig. S13). These conditions, together with the
high emission concentrations for both gas and particle phase components
(Fig. 7d–f), provide a near-ideal case study where atmospheric aging is
likely the largest factor affecting the chemical evolution of BBOA.
During this SRCF case study period, CO, NOy, and PAN mixing ratios
observed at MBO exhibited similar trends that varied dynamically and
correlated well with the fresh BBOA-1 factor (Fig. 7d–f). In addition, OA
was overwhelmingly dominated by
BBOAs, which summed to contribute 80–99 % of total OA mass (Fig. 7g).
The chemical parameters of OA and the fractional contributions of each BBOA
factor appear to be related to ∑SR (Fig. 6g and h). In order to
investigate the chemical evolution of BBOA, we reconstructed the time series
and the chemistry parameters of total BBOA
(= BBOA-1 + BBOA-2 + BBOA-3) from the residual matrix of organic
aerosol after subtracting the contributions from BL-OOA and LV-OOA. The
carbon oxidation state (OSc=2× O / C - H / C; Kroll et al., 2011) of total BBOA showed
a clear increasing trend with respect to ∑SR, consistent with the
trends of O / C and f44, while H / C, f60, and
fm/z>100 of total BBOA showed decreasing trends with ∑SR (Fig. 8).
The relationship between f44 and f60 for total OA observed during
this case study is shown in Fig. 4. f60 decreased with increased
f44 due to aging, and the data overlapped with the aged BBOA from
controlled chamber open burning of turkey oak (Cubison et al., 2011). These
results suggest oxidation of anhydrous sugar and other BBOA components due to
photochemical aging, consistent with previous observations in the laboratory
(Grieshop et al., 2009; Hennigan et al., 2011; Ortega et al., 2013) and field
(Cubison et al., 2011; May et al., 2015). In addition, the negative
correlation between BBOA-1 and ∑SR and the positive correlations of
BBOA-2 and BBOA-3 with ∑SR (Fig. 8) corroborated our earlier assumption
that BBOA-2 and BBOA-3 represented more aged, secondary BBOA whereas BBOA-1
represented primary BBOA.
Aerosol chemistry parameters of total BBOA as a function of
cumulative solar radiation for the Salmon River Complex fire case study. The
Pearson's correlation coefficients (r) are reported.
(a) OA vs. CO during 14 August at 20:00 to 16 August at
09:00 PDT with nighttime-transported plumes illustrated as black circles and
daytime-transported plumes as red crosses. The orthogonal distance regression
(ODR) results for the two plume types are shown with the 1σ
uncertainties reported for the fit slopes (s) and intercepts (i);
(b) a comparison of the average concentrations of five OA factors
(stacked) between the nighttime- and daytime-transported plumes. The average
mass fractions of the BBOAs to total OA mass in each plume type are reported;
(c) average HRMS of total BBOA for the nighttime-transported plumes;
(d) average HRMS of total BBOA for the daytime-transported plumes;
and (e) difference BBOA HRMS between day and night plumes. The
elemental ratios of BBOA calculated with the IA method are shown in the
legends of panels (c) and (d), with those obtained using
the AA method in parenthesis.
We classify the plumes according to ∑SR and designate those as
nighttime transported if ∑SR was below 500 W m-2, and we
classify the rest as daytime transported. OA concentration and CO mixing
ratio were tightly correlated, with r2=0.88 and 0.94 for nighttime-
and daytime-transported plumes, respectively (Fig. 9a). CO has been commonly
used as a stable plume tracer to account for dilution, and the slope obtained
from orthogonal fitting between OA and CO is defined as the enhancement ratio
(i.e., ΔOA /ΔCO). Change of ΔOA /ΔCO during
plume transport indicates the influence of factors other than dilution, e.g.,
SOA formation or OA evaporation. For the SRCF case study, ΔOA /ΔCO was very similar for the day plumes and the night plumes:
0.28 ± 0.014 vs. 0.27 ± 0.005 µg m-3 ppbv-1
respectively (Fig. 9a), suggesting no net OA mass enhancement due to
photochemical aging. This is consistent with the findings of Collier et
al. (2016), which compared selected BB events from this dataset measured at
MBO to those aboard a research aircraft sampling fresher plume emissions and
found very similar OA enhancements between the fresher and more aged
emissions. However, compared to daytime plumes, OA for plumes transported
during nighttime was less oxidized (Fig. 9c and d) and was dominated by the
fresh BBOA-1 (53 %), followed by the most oxidized BBOA-3 (24 %) and
intermediately oxidized BBOA-2 (15 %; Fig. 9b). By contrast, daytime
plumes were characterized by a significant decrease in the mass fraction of
BBOA-1 (37 %) coupled with increases in the fractions of BBOA-2
(20 %) and BBOA-3 (37 %). This is corroborated by the significant
differences in chemical composition for the two types of plumes, where the
average HRMS (Fig. 9c and d) indicated that the BBOA in daytime plumes had a
higher degree of oxidation (average O / C = 0.66) than the night
plumes (O / C = 0.55). These observations together suggest that,
although net OA production was conserved with higher photochemical aging,
BBOA was chemically transformed, likely due to oxidative processing in both
gas and particles phases followed by fragmentation and
volatilization.