Available via license: CC BY 4.0
Content may be subject to copyright.
1
1
Investigating size-segregated sources of elemental composition of
2
particulate matter in the South China Sea during the 2011 Vasco
3
Cruise
4
5
Miguel Ricardo A. Hilarioa, Melliza T. Cruzb, Maria Obiminda L. Cambalizaa,b, Jeffrey S. Reidc,
6
Peng Xianc, James B. Simpasa,b, Nofel D. Lagrosasa, b, *, Sherdon Niño Y. Uy b, Steve Cliff d,
7
Yongjing Zhao d
8
a Department of Physics, Ateneo de Manila University, Quezon City, Philippines
9
b Manila Observatory, Ateneo de Manila University campus, Quezon City, Philippines
10
c Marine Meteorology Division, Naval Research Laboratory, Monterey, CA, USA
11
dAir Quality Research Center, University of California Davis, CA, USA
12
Correspondence to: Maria Obiminda L. Cambaliza (mcambaliza@ateneo.edu)
13
* Now with Center for Environmental Remote Sensing, Chiba University, Japan
14
Abstract
15
The South China Sea/West Philippine Sea (SCS/WPS) is a receptor of various natural and anthropogenic aerosol
16
species from throughout greater Asia. In combination with its archipelagic/peninsular terrain and strong Asian monsoon
17
climate, the SCS/WPS hosts one of the most complex aerosol-meteorological systems in the world. However, aside from the
18
well-known biomass burning emissions from Indonesia and Borneo, the current understanding of aerosol sources is limited-
19
especially in remote marine environments. In September 2011, a 2-week research cruise was conducted near Palawan,
20
Philippines to sample the remote SCS/WPS environment. Size-segregated aerosol data was collected using a Davis Rotating-
21
drum Unit size-cut Monitor sampler and analyzed for concentrations of 28 selected elements. Positive Matrix Factorization
22
(PMF) was performed separately on the coarse, fine, and ultrafine size ranges to determine possible sources and their
23
contributions to the total particulate matter mass. Additionally, size distribution plots, time series plots, back trajectories and
24
satellite data were used in interpreting factors. Using tracers of various sources, a linear regression analysis and correlation
25
matrices showed the presence of soil dust and sea spray in the coarse mode, biomass burning in the fine mode and oil
26
combustion in the ultrafine mode. Mass distributions showed elevated aerosol concentrations towards the end of the sampling
27
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
2
period which coincided with a shift of air mass back trajectories to Southern Kalimantan. Covariance between coarse and fine
28
mode sources were observed. The PMF analysis resolved five sources across the three size ranges: biomass burning, oil
29
combustion, soil dust, sea spray and a fly ash factor largely composed of heavy metals. The agreement between the PMF and
30
the linear regression analyses suggests the robustness of the PMF solution. While biomass burning is indeed a key source of
31
aerosol, the study shows the presence of other important sources in the SCS/WPS. Understanding these sources is key to
32
characterizing the chemical profile of the SCS/WPS and, by extension, developing our understanding of aerosol-cloud behavior
33
in the region.
34
1. Introduction
35
The South China Sea/West Philippine Sea (SCS/WPS) is a receptor of a multitude of natural and anthropogenic
36
aerosol species. At the same time, due to its archipelagic/peninsular terrain coupled with a strong Asian monsoon climate, the
37
region exhibits some of the world’s most complicated meteorology. Together, the SCS/WPS hosts one of the world’s most
38
complex and sensitive composition and climate regimes (Balasubramanian et al., 2003; Yusef and Francisco, 2009; Atwood
39
et al., 2013; Reid et al., 2012, 2013, 2015). Particles in the atmosphere are known to influence radiative forcing via absorption
40
and scattering of solar radiation (Nakajima et al., 2007; Boucher et al., 2013; Lin et al., 2013; Ge et al., 2014) and act as cloud
41
condensation nuclei (CCN), affecting cloud reflectivity, evaporation and precipitation rates (Sorooshian et al., 2009; Lee et
42
al., 2012; Boucher et al., 2013; Ross et al., 2018). The northern portion of the SCS/WPS has been known to be impacted by
43
not only China via dust storms (Wang et al., 2011; Atwood et al., 2012) and industrial pollution but also by Southeast Asia
44
through its own anthropogenic pollution and biomass burning (Lin et al., 2007; Cohen et al., 2010a, b; Wang et al., 2011; Reid
45
et al., 2015, 2016). Countries surrounding the Maritime Continent (MC) are known to be impacted by strong seasonal burning
46
(Balasubramanian et al., 2003; Reid et al., 2013). The atmospheric residence times of fine particles allow for long range
47
transport, potentially creating regional and global concerns (Cohen et al., 2010a).
48
Highlighting the unique combination of terrain and sea that feeds into the complexity of the meteorological
49
environment of the region, Reid et al. (2012) and Xian et al. (2013) posed the long-range hypothesis that monsoonal flows and
50
higher-frequency meteorological phenomena are a major factor in seasonal aerosol dispersion. Biomass burning plumes are
51
known to cause severe haze episodes due to these monsoonal flows, raising concentrations of particulate matter (PM) to impact
52
cloud physics and, in some cases, to dangerous air quality levels across large areas, particularly in association with positive
53
phases of the El Niño-Southern Oscillation (ENSO) (Engling et al., 2014; Fujii et al., 2015). Likewise, biomass burning is a
54
significant contributor to the region’s CCN budget in all years as are the region’s significant anthropogenic emissions
55
(Balasubramanian et al., 2003; Field et al., 2008; Reid et al., 2012; 2013; 2015; 2016; Atwood et al., 2017).
56
Partly due to the emphasis on dramatic biomass burning as the primary source of aerosol particles in the region, the
57
contributions of other regional sources are not well understood or perhaps underappreciated. As the SCS/WPS is host to major
58
population centres, industry, major ports, and coal and oil combustion are expected to be an important regional source of
59
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
3
aerosol particles in the MC. Soil dust and coarse mode biological particles may also play a role in as ice nuclei (O’Sullivan et
60
al., 2014), as biomass burning plumes are known to entrain such particles (Reid et al., 1998; 2005; Schlosser et al., 2017). As
61
such, a network of interacting sources exists in the region surrounding the SCS/WPS, wherein sources mix during transport
62
and complicate source apportionment. Understanding the nature of sources in the remote MC and their contributions is key to
63
characterizing the aerosol environment in the SCS/WPS and its relationship with cloud behavior and precipitation patterns in
64
the region; this is particularly true given the higher sensitivity of clouds to particle perturbations at lower concentrations.
65
However, the identification of sources is complicated by their complex chemistry and interactions with the marine environment
66
(Atwood et al., 2012; 2017).
67
As part of the Seven South East Asian Studies program (7-SEAS), a research cruise (Reid et al., 2015) was conducted
68
in late September 2011 onboard the Philippine-flagged M/Y Vasco in the vicinity of the northern Palawan archipelago. The
69
goal of this cruise was to observe the behavior of aerosol particles in the SCS/WPS and test the transport hypothesis proposed
70
in Reid et al. (2012) that the Philippines is a long-range receptor of aerosol species transported across the SCS/WPS during
71
the boreal summer southwest monsoon from Borneo, Sumatra, and the Malay Peninsula. In particular, the cruise aimed to
72
observe that MC emissions were reaching the Southwest Monsoon monsoonal trough. The Palawan archipelago is a good
73
receptor site for regional emissions due to its largely rural settlements and its location upwind relative to the rest of the
74
Philippines. The sampling period coincided with the passage of one tropical storm and two tropical cyclones (TC). Of particular
75
importance is the passage of a supertyphoon Nesat beginning on 26 September as TC inflow arms are known to cause abrupt
76
changes in regional flows.
77
As part of the 2011 Vasco cruise particulate matter was collected using a size segregated Davis-Rotating Uniform
78
Size-Cut Monitor (DRUM) impactor analyzed for elemental composition. While Reid et al. (2015) noted the presence of
79
plumes in two episodes during the cruise, their initial analysis of the region’s atmospheric chemistry also suggested the events
80
were a mix of biomass burning and oil or shipping emissions due to elevated levels of vanadium. Additionally, differences in
81
elemental ratios, mass fractions and back trajectory origins between the two events support the presence of other sources
82
besides biomass burning. From the initial analysis of aerosol chemistry presented by Reid et al. (2015), this study aims to
83
investigate aerosol sources in the SCS/WPS and to further develop the current understanding of the effect of regional
84
meteorological phenomena on aerosol dispersion. The paper shows that, though biomass burning is a major source of aerosols
85
in the SCS, anthropogenic sources such as oil combustion also play an important role in the chemical profile of the region. As
86
we report, soil transport was observed as well.
87
In this paper we expand on the original 2011 Vasco cruise analysis to quantitatively apportion sampled biomass
88
burning and anthropogenic aerosol species. Positive Matrix Factorization (PMF) was performed on size-segregated PM to
89
detect possible size-specific sources (Han et al., 2006; van Pinxteren et al., 2016). Indeed, the relationship between the
90
aerodynamic diameter of a particle and its source has been well-established in literature (Reid et al., 1993; Balasubramanian
91
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
4
et al., 2003; Han et al., 2006; Lestari et al., 2009; Wimolwattanapun et al., 2010; Santoso et al., 2010; Karanisiou et al., 2009;
92
Seneviratne et al., 2010; Atwood et al., 2012; Lin et al., 2015; Cahill et al., 2016). Aerosol factors and characteristics were
93
then used to spawn back trajectories to identify individual island emissions areas.
94
2. Sampling and Methods
95
2.1. Overall cruise sampling and environment
96
A general overview of the 2011 cruise can be found in Reid et al. (2015) and a brief summary is provided here.
97
Sampling was conducted around the Palawan archipelago, an island chain located at the southwestern edge of the Philippines
98
in between the SCS/WPS and the Sulu Sea. Sampling was performed between Manila and the northern tip of Palawan Island
99
onboard the M/Y Vasco which left Manila Bay on 17 September 2011 and returned on 30 September 2011 (Fig. 1). Majority
100
of samples were collected around the areas of El Nido and Malampaya Sound (111.1° N, 119.3° E) where the vessel was on
101
station from 21-28 Sept. The largely rural population of Palawan made it an ideal receptor for regional rather than local
102
emissions.
103
104
Figure 1. Path taken by the M/Y Vasco for 17-20 September (red), 20-28 September (black), 28-30 Sept (blue). Majority
105
of sampling was done at the northern end of Palawan island. Image courtesy of GoogleMaps.
106
The cruise was conducted at the end of the boreal summer monsoon which usually lasts from June through September
107
(Loo et al., 2014; Chang et al., 2005). The Asian monsoon is caused by the annual march of the sun and asymmetrical heating
108
of air masses due to the complex terrain of Southeast Asia (Chang et al., 2005). The campaign coincided with the peak burning
109
season in Southern Kalimantan and Southern Sumatra, which have been measured to be the highest emitters of biomass burning
110
plumes in the MC (Reid et al., 2012). As the southwest monsoon is characterized by winds travelling southwest to northeast,
111
Reid et al. (2015) proposed that the Philippines was an excellent receptor for regional emissions from the MC.
112
Although 2011 was a moderate La-Niña year, it was noted that fire activity and precipitation levels resembled a
113
neutral year (Reid et al., 2015). The cruise took place when the Madden-Julian Oscillation (MJO) was transitioning from the
114
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
5
wet phase to the dry phase, which is expected to enhance burning activity and transport. With the passage of tropical cyclones
115
(TCs), significant aerosol events were observed to propagate across the region.
116
Reid et al. (2015) described three tropical events that occurred during the cruise, specifically tropical storm (TS)
117
Haitang, super-TC Nesat, and super-TC Nalgae. The presence of inflow arms in the SCS has been suggested to affect the
118
aerosol environment by bringing more MC air into the region (Reid et al., 2015). The passage of Nesat was observed to abruptly
119
affect air mass trajectories coinciding with an enhancement of several elements during the last two days of the cruise.
120
Figure 2 shows the evolution of the meteorological environment over the cruise period with comparisons between
121
satellite-derived aerosol optical depth (AOD) derived from the MODIS-Terra and MODIS-Aqua satellites, back trajectories
122
from NOAA Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) and 850 hPa smoke concentrations
123
from the Navy Aerosol Analysis and Prediction System (NAAPS).
124
125
Figure 2. Satellite images of the SCS/WPS region taken from (a-d) NASA Worldview with overlayed AOD, (e-h)
126
NAAPS smoke concentration plots (μg/m3; 850 hPa) and (i-l) HYSPLIT ensemble back trajectories during the cruise
127
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
6
(isobaric, 300m AGL, 72 hours, ending at 00:00 UTC/08:00 LST) for 18, 22, 26 and 29 Sept. Red star indicates location
128
of the Vasco.
129
2.2. Aerosol sampling and analysis
130
Size-resolved aerosol samples were collected during the cruise using a Davis-Rotating Unit for Monitoring (DRUM)
131
continuously sampling cascade impactor. Samples were collected with a 10 μm inlet and eight size cuts at 5, 2.5, 1.15, 0.75,
132
0.56, 0.34, 0.26, 0.07 μm at a 90-minute time resolution from noontime 17 September until noontime 30 September local-time.
133
Particles were collected on Mylar strips coated with Apiezon grease. The eight drums were rotated at a consistent rate to create
134
a temporal record of mass concentration (Raabe et al., 1988). X-ray fluorescence (XRF) was performed on the DRUM samples
135
at the Advanced Light Source (ALS) of Lawrence Berkeley National Laboratory to measure mass concentrations of 28
136
elements ranging from Na to Pb. In this study, data was filtered based on location notes from the cruise such that samples
137
collected in the vicinity of Manila Bay were excluded from the analysis. Additionally, samples during an 8-hour pump failure
138
that occurred on 20 September were also excluded from the dataset. In the analysis, the stages were aggregated into three
139
modes: coarse (1.15-10 μm), fine (0.34-1.15 μm) and ultrafine (0.07-0.34 μm) modes.
140
2.3. Model and satellite data
141
NOAA Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) back trajectories (Draxler et al.,
142
1998, 1999) were generated throughout the cruise period to investigate locations of aerosol emission. HYSPLIT back
143
trajectories have been used in several studies to establish air mass source regions (Lin et al., 2007; Cohen et al., 2010a; Atwood
144
et al. 2012, 2017). Back trajectories were run for 72 hours for heights of 500 m and 300 m to investigate possible vertical
145
inhomogeneity that has been noted in other SCS/WPS papers (Atwood et al., 2012). Trajectory endpoints corresponded to
146
cruise coordinates. Trajectories were constrained isobarically to limit vertical wind velocity since our area of interest is surface-
147
level emission.
148
The Navy Aerosol Analysis and Prediction System (NAAPS) reanalysis product (Lynch et al., 2016) with driving
149
meteorology was used to provide overall aerosol and meteorological context to the analysis. This reanalysis utilizes a modified
150
version of the NAAPS as its core and assimilates quality controlled retrievals of aerosol optical depth (AOD) from Moderate
151
Resolution Imaging Spectroradiometer (MODIS) on Terra and Aqua and the Multi-angle Imaging SpectroRadiometer (MISR)
152
on Terra (Zhang et al., 2006; Hyer et al., 2011; Shi et al., 2014). NAAPS characterizes anthropogenic and biogenic fine
153
(including sulfate, and primary and secondary organic aerosols), dust, biomass burning smoke and sea salt aerosols. Smoke
154
from biomass burning is derived from near-real time satellite based thermal anomaly data to construct smoke source functions
155
(Reid et al., 2009), with additional orbital corrections on MODIS based emissions and regional tunings. The system has been
156
successfully used to monitor biomass burning plumes and to study the relationship of aerosol lifecycle to weather systems over
157
the MC (Reid et al., 2012, 2015, 2016; Atwood et al., 2013; Xian et al., 2013).
158
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
7
Active fire hotspot data was downloaded from the Fire Information for Resource Management System (FIRMS)
159
(https://firms.modaps.eosdis.nasa.gov/). Active fire hotspots and aerosol optical depth (AOD) at a wavelength of 550 nm were
160
tracked throughout the cruise via the Moderate Resolution Imaging Spectroradiometer (MODIS). MODIS detects thermal
161
anomalies across a region to identify possible fire activity. MODIS-derived AOD was used to derive large-scale estimates of
162
PM2.5 in some studies (e.g., Zheng et al., 2017). In the study, MODIS was used to track burning emissions which were found
163
to be particularly prevalent in Eastern Malaysia and Indonesia. The use of MODIS to track active fire hotspots has been used
164
in other studies to understand seasonal trends in agricultural burning (Reid et al., 2012) and to identify and locate burning-
165
related sources when used in conjunction with HYSPLIT back trajectories (Atwood et al., 2017).
166
The NASA Worldview site (www.worldview.nasa.gov), an application operated by the NASA/Goddard Space Flight
167
Center Earth Science Data and Information System (ESDIS) project, was used to supplement the satellite data by providing
168
true color images of the region and is particularly useful in demonstrating sudden changes of cloud environment or monsoon
169
flow caused by tropical cyclones.
170
2.4. Positive Matrix Factorization
171
Positive Matrix Factorization (PMF) was used to study the covariability of elemental species. PMF is a multivariate
172
factor analysis technique used in source apportionment that resolves a sample matrix X ( ) of samples and species into
173
matrices F ( ), G ( ) and E ( ), the source contribution matrix, source profile matrix and residual matrix,
174
respectively, with the assumption of k factors:
175
176
The goal of PMF is to determine the number of factors or sources k such that the solution will be physically interpretable.
177
Developed by Paatero and Tapper (Paatero and Tapper, 1994), PMF is a well-established approach used in previous source
178
apportionment studies (Polissar et al., 1998; Lee et al., 1999; Han et al., 2006; Chan et al., 2008; Karanisiou et al., 2009; Lestari
179
et al., 2009; Santoso et al., 2010; Wimolwattanapun et al., 2010). PMF provides more physically realistic results compared to
180
other factor analysis techniques due to non-negative constraints in the model and better treatment of missing or below detection
181
limit (BDL) values by increasing the associated uncertainty (Paterson et al., 1999).
182
PMF outputs source profiles (F) and source contributions (G). PMF source profiles were normalized to the percent of
183
species sum, defined as the percent concentration of an element apportioned to a source. A robust mode of PMF was used for
184
analysis, characterized by the parameter . is the outlier threshold distance which reduces the effect of extremely large data
185
points and is set at a value of 4.0 to be consistent with other PMF studies (Lee et al., 1999; Han et al., 2006).
186
For each of the three modes, a minimum Pearson R value of 0.0 was used to filter species based on correlations with the
187
mode-aggregated elemental PM concentration. This means that species negatively correlated with the summed elemental PM
188
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
8
concentration were not included. From the 28 species identified by XRF per DRUM stage, 20 species in the coarse mode, 22
189
species in the fine mode, and 19 species in the ultrafine mode were used in the size-resolved PMF. Tables S1-3 (Supplementary
190
Information) show the correlation coefficients of the elements, with filtered elements having negative Pearson R values against
191
the summed elemental PM concentration. The filtering of species through correlation was observed to improve the
192
interpretability of the source profiles and remove the need for the matrix rotation parameter, Fpeak.
193
Data screening was performed based on the approach of Polissar et al. (1998) to ensure that no erroneous data points were
194
included in the analysis. Signal-to-noise ratios were determined and species with low ratios (less than 0.2) were excluded from
195
the data set (Paatero and Hopke, 2003). Below detection limit (BDL) values were replaced with half the detection limit (Han
196
et al., 2006). Detection limit values and error values were based on values provided by the Lawrence Berkeley National
197
Laboratory.
198
The current study employs a size-resolved PMF approach as a supplement to the other analysis methods. PMF is a
199
powerful tool that quantifies the contributions of PM sources and is useful for forming an initial understanding of the possible
200
sources from the data. However, PMF may neglect important events, particularly short-term ones, that can reveal insightful
201
interactions between identified sources and is unable to dissociate covarying sources as it assumes orthogonality between
202
factors (Van Pinxteren et al., 2016).
203
3. Results I: Mass distributions and time series of selected elements
204
3.1. DRUM mass distributions
205
Mass size distributions show normalized species concentrations (dM/dlogDp) across all eight DRUM stages and can be
206
used to validate the signal of a mode-specific tracer. In addition to isolating the signal of a tracer, changes in the mass
207
distributions of key elements over time indicate periods when mode-specific sources are present. Figure 3 depicts the mass
208
size distributions of key elements (a) potassium (K) as a tracer for biomass burning in the fine and ultrafine modes, (b) sulfur
209
(S), which is used as an indicator of combustion, (c) aluminum (Al) and (d) silicon (Si) which are often paired as tracers for
210
soil dust, (e) vanadium (V) and (f) nickel (Ni), which are often paired as tracers of oil combustion, (g) iron (Fe), another key
211
tracer for dust, and (h) chlorine (Cl), a reasonable tracer for sea spray given the sampling location. Figure 3 is further divided
212
into time periods, distinguished by color: 18-19 September (red), 19-24 September (blue), 24-27 September (green) and 27-
213
30 September (black).
214
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
9
215
Figure 3. Time evolution of mass size distributions of key elements over the cruise period. (a) potassium, (b) sulfur, (c)
216
aluminum, (d) silicon, (e) vanadium, (f) nickel, (g) iron, and (h) chlorine. Time periods are colored: 18-19 Sept (red),
217
19-24 Sept (blue), 24-27 Sept (green), 27-30 Sept (black). Stage numbers are depicted in (a).
218
219
Elements associated with combustion showed generally bimodal distributions with stage 5 (0.56-0.75 μm) and stage 7
220
(0.26-0.34 μm) peaks. K, S, Al, and Si have very similar mass size distributions over the cruise period which are suggestive of
221
a common source (Fig. 3a-d). These elements have strong peaks in stage 5 and 7 during the whole cruise but particularly high
222
values are observed during the last days of the sampling period (27-30 Sept). A general enhancement late in the cruise is likely
223
related to the increase in the number of active fire hotspots reported by Reid et al. (2015), who attributed these hotspots
224
primarily to Indonesian Kalimantan and Southern Sumatra. As the cruise took place during the end of the boreal summer, 300
225
m a.g.l. winds were predominantly southwesterly. A shift in back trajectories at the end of the cruise to the western and southern
226
coasts of Borneo is observable in Fig. 2l, suggesting the source of the late-cruise enhancement to be the MC, which hosts
227
elevated aerosol background levels from seasonal burning (Reid et al., 2013). The advection of this large aerosol event can be
228
observed in the NAAPS smoke model over the region (Fig. 2g, h). The attribution of late-cruise aerosol enhancement to the
229
6
1
2
4
5
7
8
3
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
10
MC is in agreement with Reid et al. (2015) who noted that the AOD maps and southwesterly flows towards the end of the
230
cruise were suggestive of southwesterly transport from the MC to SCS/WPS.
231
Covarying behaviors of Al and Si with K and S suggest possible fine soil entrainment caught in burning updraft (Reid et
232
al., 2015). The stage 5 and stage 7 peaks in S are similar to those observed for northern SCS/WPS in the springtime (Atwood
233
et al., 2012); however, we report enhanced values, attributed to the timing of the sampling period during the MC burning
234
season.
235
Interestingly, Si shows a strong peak early in the cruise (18-19 Sept) unique to the ultrafine mode which indicates this
236
particular signal may not originate from soil dust but fly ash (Xie et al., 2009). As the Vasco was travelling past the islands of
237
Mindoro and Coron en route to Palawan, local sources are likely the cause of the ultrafine Si enhancement. This early-cruise
238
Si signal is further examined through later time series and regressions.
239
V shows a mass distribution characteristic of a combustion source with strong peaks in stage 5, stage 7, and stage 8 (0.07-
240
0.26 μm) (Fig. 3e). Almost no contribution was observed for coarser stages 1 through 4 (0.75 -10 μm), indicating that V did
241
not originate from soil (Lin et al., 2015) and can be treated as a tracer for oil combustion. Ni shows a similar bimodal mass
242
distribution (Fig. 3f) but had a larger spread over the eight stages than V, which may be due to contributions from other sources
243
such as fly ash (Davison et al., 1974).
244
Fe and Cl, well-known tracers for soil dust and sea spray, respectively, showed coarse-mode distributions that taper off
245
considerably in the submicron stages (Fig. 3g, h). Cl shows a purely coarse distribution which suggests sea spray considering
246
the sample location (Viana et al., 2008; Gugamsetty et al., 2012; Farao et al., 2014). Fe shows small peaks in stage 4 (0.75-
247
1.15 μm), stage 5, and stage 7; however, these do not constitute a significant signal relative to its coarse mode concentrations.
248
As such, we treat Fe as our coarse mode soil dust tracer. The mass distribution of Fe is observed to increase across stages 1
249
through 3 (2.5-10 μm) over the cruise period. The increase in coarse Fe coincides with the NAAPS-simulated transport of
250
smoke (Fig. 2g, h) and mirrors the enhancements of K, S, Si (Fig. 4a, b), and Al (Fig. S1a, b). These patterns suggest that
251
coarse soil dust accompanies smoke emissions, possibly through entrainment. The presence of soil dust is further corroborated
252
by Fig. 3c and Fig. 3d, which show the presence of Al and Si in the coarse mode. The distinct coarse and fine mode peaks of
253
Al and Si indicate separate soil dust sources. As fine mode particles have longer residence times (Cohen et al., 2010a), the fine
254
peaks may be an indicator of long-range transport of fine soil dust through the SCS/WPS.
255
Interpreting DRUM data reveals insights about the composition and interpretation of sources. Table 1 shows the ratios
256
of elemental PM1.15/PM10 mass concentrations. As in Atwood et al. (2012), the ratio-slope was computed by taking the slope
257
of the linear regression line between PM1.15 and PM10 mass concentrations, accompanied by r2 values. Direct averages of per-
258
timestamp ratios of PM1.15 and PM10 were also taken to compute for ratio-averages, accompanied by the standard deviation of
259
the ratios. Fe and Cl both had ratios of 0.06, which confirm the predominantly coarse nature of these species. As commonly
260
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
11
used tracers of soil dust, Al and Si show moderate ratio-slope values of 0.51 and 0.29, respectively, suggesting that Al resides
261
in both coarse and fine (PM1.15) modes while Si is predominantly coarse. As expected, elements commonly associated with
262
anthropogenic species such as V, K, and S show high PM1.15/PM10 ratios (0.8 and above) which indicate that these elemental
263
particles largely reside in the fine and ultrafine modes. The high ratios of V, K, and S provide evidence for the presence of
264
anthropogenic emissions from sources such as oil combustion and biomass burning while the low ratios of Fe and Cl support
265
their treatment as tracers for soil dust and sea spray, respectively.
266
The time-resolved DRUM data is important for showing variations in species which may be representative of
267
important aerosol events. Thus, observations on the time-resolved DRUM data can aid in our analysis. At the beginning of the
268
cruise, between 18 to 19 Sept, V, Ni, and Si show enhancements in stages 5, 7, and 8. The stage 7 Si peak during this time is
269
the maximum concentration over the entire cruise period, so this warrants further analysis through later time series and
270
regressions. The period of 19-24 Sept shows a low point in the DRUM peaks of several elements, most notably combustion
271
tracers K and V (Fig. 3a, c), while Cl (Fig. 3h) shows higher peaks in the coarse-mode which suggests a period of clean marine
272
aerosol. This period was described by Reid et al. (2015) as the cleanest of the cruise. The NAAPS model shows nearly zero
273
smoke concentration at the sampling site (Fig. 2f) while 72-h HYSPLIT back trajectories indicate that air masses originate
274
from central SCS/WPS (Fig. 2j). From 24 to 27 Sept, we observed the first major aerosol event characterized by the stage 5
275
and 7 enhancements of several combustion elements: K, S, Al, Si, V, and Ni (Fig. 3a-f). Fe, our coarse-mode soil dust tracer,
276
shows enhancements in stages 1 to 3 (Fig. 3g), which points to combustion-related entrainment of soil dust in the coarse mode.
277
The NAAPS model (Fig. 2g) depicts the intensification and spread of a smoke-related aerosol event that had been escalating
278
in southern Kalimantan since 22 Sept, reaching the Vasco around 26 Sept. During this mid-cruise period, concentrations of
279
biomass burning species K, S, Si, Al are elevated, and oil combustion tracers V and Ni show their maximum concentrations
280
for the cruise in stages 5 and 7 (Fig. 4e, f). The last period, 28 to 30 Sept, depicts the highest concentrations of elements
281
associated with biomass burning (Fig. 4a, b; Fig. S1a, b). As seen in the NAAPS smoke model (Fig. 2h) and HYSPLIT model
282
(Fig. 2l), the westward movement of TC Nesat across the region alters back trajectories to wind around Borneo island, reaching
283
southern Kalimantan which hosted a high active fire hotspot density during the time (Reid et al., 2015), thus bringing polluted
284
air masses toward the sampling site. Stage 5 and 7 peaks of K and S are quite notable as no other stages show significant
285
enhancements in response to this event. Fe, Al, and Si show similar changes but for the coarser stages 1 to 3 (Fig. 3g), indicating
286
a covariance of soil dust and biomass burning tracers. The temporal trends from the DRUM data serve as an entry point into
287
the time series analysis. By identifying key DRUM stages and time periods per element based on their mass size distributions,
288
we can then examine these stages to observe aerosol events over the cruise period.
289
3.2. Time series of selected elements
290
291
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
12
Figure 4. Time series of (a) Stage 4-6 K, S, Si, (b) Stage 7-8 K, S, Si, (c) Stage 4-6 V, Ni, (d) Stage 7-8 V, Ni, and (e) Stage
292
1-3 Fe, Cl.
293
The first few days of the cruise showed an 18 Sept event in oil combustion tracers V and Ni in the ultrafine mode (Fig.
294
4d) with a coincident but lower-magnitude response in the fine mode (Fig. 4c). Ultrafine mode V and Ni show their maxima
295
for the cruise period during this time, expanded further in Section 5. High concentrations of ultrafine Si were sampled during
296
this time from the beginning of the cruise until 19 Sept when it dropped to stable background levels. This early-cruise
297
enhancement was also seen in its mass distribution plot (Fig. 3c). As the Vasco was traveling among islands, the Si signal may
298
be due to local sources en route to the El Nido sampling site.
299
Reid et al. (2015) noted periods of clean regime after departing Manila Bay through midday 22 Sept, observable in the
300
consistently low concentrations of various elements (Fig. 4). Chlorine shows a gradual increase in concentration from 20 Sept
301
until 24 Sept. Chlorine, although it ages into HCl, is assumed to be fresh due to the sampling location and can therefore be
302
used as an indicator of sea spray. Interestingly, coarse-mode Cl (Fig. 4e) showed peak concentration times during low points
303
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
13
in the concentrations of anthropogenic aerosol species (Fig. 4a-d), marking periods of clean marine aerosol on 22-24 Sept and
304
26-28 Sept, likely through wet deposition processes. During these times, back trajectories shift away from source regions and
305
traverse open sea (Fig. 2j, k) which also hosts a lower shipping route density compared to coastal regions (Fig. S2,
306
Supplementary Information). The first half of the cruise also saw the lowest concentrations from species associated with
307
biomass burning, specifically submicron K, S, Si, (Fig. 4a, b), and Al (Fig. S1a, b, Supplementary Information). These species
308
track each other quite well throughout the cruise period indicating a common source.
309
The event between 24 Sept and 26 Sept is observable on the time series of several key elements. The plume was the first
310
of two distinct plume events reported by Reid et al. (2015) with the later plume occurring on 29 Sept. The enhancement of all
311
elements in Fig. 4 suggests a mix of biomass burning, oil combustion and soil dust influences within the 24-26 Sept plume.
312
Fine mode V and Ni show their maximum concentrations for the cruise during this event (Fig. 4c). Although these two plumes
313
appeared as one uniform progression across the SCS/WPS region on the NAAPS smoke model (Fig. 2h), the time series
314
showed the presence of two distinct events (Fig. 4), which is corroborated by observations from Reid et al. (2015). During this
315
period, plume concentration dropped sharply before recovering due to the passage of squall lines sharp, observed in the time
316
series for K, S, Si, Fe, and Cl (Fig. 4a, b, e). As concluded in Reid et al. (2015), frequent, short-term events such as this must
317
be accounted for in studies on aerosol-convection interaction.
318
The period between plumes (26-28 Sept) is characterized by an overall drop in the aerosol concentration of species
319
associated with anthropogenic sources (K, S, V, Ni; Fig. 4a-d). As Cl concentrations show peak values during this period (Fig.
320
4d), this indicates a period of pure marine aerosol sampling similar to the 22-24 Sept clean period. Coinciding with the passage
321
of TC Nesat through the SCS/WPS, the observed drop in aerosol concentration is attributed to a possible restriction of shipping
322
traffic in response to the TC and scavenging of aerosols by precipitation along the TC inflow arm (Fig. 2c) (Reid et al., 2015).
323
The last days of the cruise were particularly eventful as the largest aerosol event of the cruise period was visible on the
324
NAAPS model in the form of smoke (Fig. 2h), accompanied by the spread of high AOD values throughout the SCS/WPS (Fig.
325
2d). Although the large areas of cloud cover created by TC Nesat hinders the detection of AOD on 26 Sept, the region is free
326
of cloud cover by 29 Sept that significant AOD values were observed to visibly stretch from Southern Kalimantan towards the
327
Vasco sampling site (Fig. 2d). In general, the NAAPS smoke transport model agrees with the spatial distribution of high AOD.
328
Here, NAAPS modelling of smoke transport is useful in demonstrating the event’s northward advection and the severity of
329
smoke concentration in Borneo island on 26 Sept (Fig. 2h). Time series plots of elements associated with biomass burning (K,
330
S, Si; Fig. 4a, b) and coarse mode soil dust (Fe; Fig. 4d) show significant enhancements during this time which were also
331
observed on their mass distributions (Fig. 3). HYSPLIT back trajectories show that air masses originate from Southern
332
Kalimantan during this period as opposed to mainland Malaysia during the first half (Fig. 2j, l). The shift in air mass trajectories
333
is attributed to the passage of TC Nesat through the region as inflow arms from TCs have been observed to accelerate air mass
334
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
14
advection across the SCS/WPS, bringing more MC air into the region (Reid et al., 2012; 2015). The observed transport of
335
emissions from Borneo indicates that TC-enhanced long-range transport is a significant factor in SCS/WPS aerosol dispersion.
336
4. Results II: positive matrix factorization and regressions
337
4.1. Source apportionment via positive matrix factorization
338
To verify groupings of key elements and aid in source identification, size-resolved PMF was performed. As described in
339
Section 2, the eight-stage DRUM data were combined into coarse (1.15-10 μm), fine (0.34-1.15 μm) and ultrafine (0.07-0.34
340
μm) modes and the species included in the PMF analysis were then filtered based on their correlation to the aggregated PM
341
concentration. The PMF analysis resolved five sources across the three size ranges: biomass burning, oil combustion, soil dust,
342
sea spray and fly ash (Table 2).
343
Figure 5 shows the percent contribution of each source relative to the total elemental PM mass. One strength of PMF is
344
its quantification of a source’s contribution. As expected, natural sources such as the crustal source/sea spray and soil dust
345
mainly contribute to the coarse mode while combustion-related sources such as biomass burning and oil combustion exist in
346
the fine and ultrafine modes. The existence of these sources in their expected modes is an indicator of the successful
347
implementation of PMF. The following sections describe the observed characteristics of sources determined by PMF.
348
349
350
Figure 5. Contributions of factors to the total elemental PM mass.
351
Sea
Spray
75%
Soil
Dust
25%
(a) Coarse (1.15 - 10 μm)
Sea
Spray
31%
Soil
Dust
12%
Biomass
Burning
34%
Oil
Combustion
9%
Fly Ash
14%
(b) Fine (0.34 - 1.15 μm)
Biomass
Burning
43%
Oil Combustion
29%
Fly Ash
28%
(c) Ultrafine (0.10 - 0.34 μm)
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
15
352
353
Figure 6. PMF source profiles across different size ranges displayed by percent of species sum for (a) sea spray, (b) soil
354
dust, (c) biomass burning, (d) oil combustion, and (e) fly ash. Coarse: Stage 1-3 (1.15-10 μm; blue), Fine: Stage 4-6
355
(0.34-1.15 μm; orange), Ultrafine: Stage 7-8 (0.07-0.34 μm; black).
356
Sea spray: This factor was resolved in the coarse and fine modes characterized by strong apportionments for Na, Mg,
357
Cl, P, and S in the coarse mode, and Cl and Mg in the fine mode as shown in the source profile of Fig. 6a. These elements are
358
indicative of sea spray (Han et al., 2006; Wang et al., 2014). Cl has been treated as a sea spray tracer under the assumption
359
that the sampled Cl originated from freshly produced sea spray (Atwood et al., 2012). This is likely the case for the cruise as
360
sampling was done over sea water. The factor showed quite high mass contributions to the coarse (75%) and fine (31%) modes,
361
attributed to the sampling location over water (Fig. 5a, b).
362
0
50
100
NaMg Al Si P S Cl KCa Ti VCr Mn Fe Co Ni Cu Zn Ga As Se Br Rb Sr YZr Mo Pb
% of Species Sum
Fly Ash
0
50
100
Na Mg Al Si P S Cl KCa Ti VCr Mn Fe Co Ni Cu Zn Ga As Se Br Rb Sr YZr Mo Pb
% of Species Sum
Soil Dust
0
50
100
NaMg Al Si P S Cl KCa Ti VCr Mn Fe Co Ni Cu Zn Ga As Se Br Rb Sr YZr Mo Pb
% of Species Sum
Sea Spray
0
50
100
NaMg Al Si P S Cl KCa Ti VCr Mn Fe Co Ni Cu Zn Ga As Se Br Rb Sr YZr Mo Pb
% of Species Sum
Biomass Burning
0
50
100
NaMg Al Si P S Cl KCa Ti VCr Mn Fe Co Ni Cu Zn Ga As Se Br Rb Sr YZr Mo Pb
% of Species Sum
Oil Combustion
(a)
(b)
(e)
(c)
(d)
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
16
Soil dust: This factor was characterized by the presence of Fe, Al, Si, K, Ca, Ti, and Zn in the coarse mode and Fe, Cr,
363
Mn, and Y in the fine mode (Fig. 6b; Table 2). Several of these elements are associated with soil dust (Artaxo et al., 1990,
364
1998; Lestari et al., 2009; Wimolwattanapun et al., 2010; Gugamsetty et al., 2012). Soil dust may originate from the nearby
365
island of Palawan but also can potentially come from Borneo. The PMF model was able to distinguish between the sea spray
366
and soil dust factors. As sea spray aerosol is assumed to be freshly sampled during the cruise and the temporal trends of the
367
two sources are distinct (Fig. 7a, b), this suggests the possibility of a long-range transport mechanism for coarse mode soil
368
dust. Fe serves as our tracer for soil dust due to its high apportionment in both soil dust modes. This factor showed mass
369
contributions of 25% and 12% in the coarse and fine modes, respectively, which indicates the predominantly coarse mode
370
contribution of the factor (Fig. 5a, b).
371
Biomass burning: This factor was characterized by high levels of K and S, and moderate levels of Al, As, and Si which
372
were found to be associated with biomass burning in previous studies (Artaxo et al., 1998; Han et al., 2006; Lestari et al., 2009;
373
Atwood et al., 2012; Alam et al., 2014) (Fig. 6c; Table 2). The factor showed the highest percent contributions to the PM mass:
374
34% and 43% in the fine and ultrafine modes, respectively. The sources of the 26 Sept and 28-30 Sept events (Fig. 7c) will be
375
investigated in Section 5. The presence of crustal elements Fe, Si, and Al in the source profile and the covariance of the coarse
376
soil dust factor (Fig. 7b) with this factor (Fig. 7c) indicate possible soil dust entrainment during burning updraft (Reid et al.,
377
2015; Schlosser et al., 2017).
378
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
17
379
Figure 7. PMF source contributions across size ranges displayed concentration (ng/m3) for (a) sea spray/crustal source,
380
(b) soil dust, (c) biomass burning, (d) oil combustion, and (e) fly ash.
381
Oil combustion: This factor was characterized by high levels of V (Fig. 7d; Table 2), a well-documented tracer for oil
382
combustion (Hedberg et al., 2005; Mazzei et al., 2008; Becagli et al., 2012). As shown in Fig. 5, the oil combustion factor only
383
appeared in the fine and ultrafine sizes, contributing 9% and 29%, respectively, to the total elemental PM mass. The increasing
384
contribution towards finer stages corroborates the identification of the factor as an anthropogenic source. The presence of oil
385
combustion is expected as the SCS/WPS hosts high shipping volume, particularly in parts of the Borneo coast (Fig. S2).
386
Fly ash: This factor was observed in the fine and ultrafine modes, characterized by high levels of trace metals Ti, Ni, Zn,
387
Se, Br, Rb, Y, and Pb in the fine mode; Fe, Ni, Zn, As, Se, Br, Rb, and Pb in the ultrafine mode (Fig. 6e); and a source
388
0
1000
2000
0
5000
10000
15000
Fine (ng m-3)
Coarse (ng m-3)
Sea Spray
(a)
0
200
400
600
0
5000
10000
Fine (ng m-3)
Coarse (ng m-3)
Soil Dust
(b)
0
2000
4000
0
500
1000
1500
Fine (ng m-3)
Ultrafine (ng m-3)
Biomass Burning
(c)
0
1000
2000
0
500
1000
Fine (ng m-3)
Ultrafine (ng m-3)
Oil Combustion
(d)
0
100
200
300
400
0
100
200
300
09/18 09/20 09/22 09/24 09/26 09/28 09/30
Fine (ng m-3)
Ultrafine (ng m-3)
Fly Ash
(e)
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
18
contribution without distinct events (Fig. 7e). The presence of As, Se, Zn, and Ni are indicative of fly ash (Davison et al., 1974;
389
Markowski et al., 1985; Deonarine et al., 2015). The source contribution time series shows a background-type signal. The
390
factor contributed 14% and 28% to the total elemental PM mass for the fine and ultrafine size ranges, respectively (Fig. 5),
391
which is indicative of a combustion-type source. Long-range transport of fly ash from coal-fired power plants in Indonesia or
392
mainland Malaysia may be responsible for the appearance of the factor as no local coal-fired power plants could be found
393
upwind of the sampling site in 2011.
394
The PMF analysis resolved the presence of five sources across the ultrafine, fine and coarse modes which aids in directing
395
further analysis by identifying key species in the source profiles. Pearson correlation heatmaps (Fig. S3-5, Supplementary
396
Information) and matrices (Tables S1-S3, Supplementary Information) were constructed to examine the relationships between
397
species. The first column of the correlation outputs (Fig. S3-5, Tables S1-S3, Supplementary Information) shows the
398
correlation coefficient of the element when compared to the summed elemental PM for that mode. Similar groupings of
399
elements were observed when compared to the PMF source profiles, indicating the robustness of the analysis. In the coarse
400
mode (Fig. S3; Table S1, Supplementary Information), we observe high correlations between Na, Mg, Cl, P, S, K, Ca, Br, and
401
Sr, which are associated with sea spray (Han et al., 2006; Wang et al., 2014). Fe, Ti, Mn, Si, and Zn show moderate to high
402
correlations in the coarse mode, indicative of dust (Karanisiou et al., 2009; Wimolwattanapun et al., 2010; Lin et al., 2015;
403
Landis et al., 2017). In the fine mode, moderate to high correlations between Al, Si, P, S, K, Br are observed (Fig. S4, Table
404
S2, Supplementary Information). Several of these biomass burning elements show similarly strong correlations in the ultrafine
405
mode (Fig. S5, Table S3, Supplementary Information). V and Ni show a high correlation coefficient (0.91) in the ultrafine
406
mode, indicative of oil combustion. The excellent correspondence between the observed groupings of elements based on
407
correlation (Tables S2-4, Supplementary Information) and the sources resolved by PMF (Table 2) adds confidence to the
408
identification of key sources during the cruise. However, as PMF is an unsupervised technique, it may miss significant aerosol
409
events, particularly transient ones. To further expand on the relationships between elements, we turn to regression analysis.
410
4.2. Regressions of selected elements
411
An early-cruise ultrafine Si event was shown in the mass size distribution (Fig. 3d) and the time series (Fig. 4b) of
412
ultrafine Si. Fly ash was the hypothesized source of the ultrafine Si signal; however, although the PMF analysis showed the
413
presence of fly ash, Si was not attributed significantly to the fly ash factor (Fig. 6e). Additionally, none of the factor
414
contributions from PMF showed a similar trend between 18-19 Sept as ultrafine Si. Regressions show that, between 18 and 19
415
Sept, Si had distinct ratio slopes and the highest correlations with P (r2 = 0.76), S (r2 = 0.73), and Al (r2 = 0.61) (Fig. 8; Table
416
S4) but showed poor correlations with other fly ash elements (As, Se, Pb; r2 < 0.12). As it was early in the cruise, the Vasco
417
was travelling past nearby islands en route to Palawan. Therefore, local sources en route to Palawan may be the source of the
418
ultrafine Si enhancement.
419
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
19
420
Figure 8. Linear regressions of ultrafine Si and its most highly correlated elements (a) P, (b) S, (c) Al, divided by
421
cruise period before Sept 19 (red) and after Sept 19 (blue).
422
As S is an indicator of general combustion (Atwood et al., 2012), it is important to elucidate its relationship with
423
tracers of other combustion sources. Fine mode and ultrafine mode linear regressions of K and V, colored by the raw
424
concentration of S at each timestamp, were taken to show the relationships between the three species (Fig. 9a, b). S is seen to
425
covary more with K than V as seen with the clearer color gradient following the K-axis, suggesting the origin of S during the
426
cruise to be more dominantly from biomass burning rather than oil combustion. Multiple linear regression was also performed
427
for these elements on the fine and ultrafine modes (Fig. S6, Supplementary Information). It was found that K and V were
428
excellent predictors of S for most of the cruise but the model required the addition of Al to capture the variance in S between
429
24 and 26 Sept. A detailed description of the multiple linear regression analysis can be found in the Supplementary.
430
The ratio between V and Ni is often used as an indicator of the type of oil combustion source (Hedberg et al., 2005;
431
Nigam et al., 2006; Mazzei et al., 2008; Becagli et al., 2012; Lin et al., 2015). Linear regression plots of V and Ni have a slope
432
of 3.64 in the ultrafine mode (Fig. 9c). Nigam et al. (2006) measured a V/Ni ratio of 3.5-4 when sampling shipping emissions
433
directly from the exhausts of various ship engines which suggests shipping to be the main source of ultrafine mode oil
434
combustion during the cruise.
435
As soil composition varies geographically, soil dust ratios are excellent indicators of a plume’s origin (Prospero et
436
al., 1999; Song et al., 2006; Witt et al., 2006). Figure 9d shows linear regressions of soil dust elements in the coarse and fine
437
modes. Al and Si, well-known indicators of dust (Viana et al., 2008; Tian et al., 2016; Landis et al., 2017), show moderate
438
correlations with each other in the coarse and fine modes but slightly differ in ratio-slopes between the fine (Si/Al ~ 1.3; r2 =
439
0.94) and coarse (Si/Al ~ 0.93; r2 = 0.78) modes (Fig. 9d). This indicates a source of fine mode soil dust source enriched in Si;
440
however, this could also be a matrix effect from the XRF analysis. As the Vasco remained near Palawan island, local dust
441
could be the source of coarse-mode Si-enrichment; however, soil dust from Borneo is also a possibility.
442
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
20
443
Figure 9. Scatter plot of key species during the cruise. (a) fine mode K, V colored by the raw concentrations of S per
444
timestamp, (b) ultrafine mode K, V likewise colored by raw concentrations of S per timestamp, (c) ultrafine mode V,
445
Ni, and (d) coarse and fine mode Al, Si.
446
5. Results III: Back trajectory analysis
447
18-19 Sept: Ultrafine V, Ni enhancement from Sandakan, Sabah
448
As described in Section 3, ultrafine mode V and Ni show a maximum around 18 Sept (Fig. 4d). As the Vasco was traveling
449
near local islands, the event may originate from a local source; however, back trajectories propose an oil combustion source
450
in Borneo. Back trajectories were generated every hour between 14:00 to 18:00 UTC (corresponding to 22:00 to 02:00 LST)
451
on 18 September and show a westward shift along the eastern coast of Borneo (Fig. 10a). The coast of Borneo is largely forest
452
(Fig. 10b) but hosts the city of Sandakan, one of Sabah’s major ports (Fig. 10c, d). In addition to shipping traffic (Fig. 10d),
453
Sandakan contains oil depots which are a major source of industry in the area. During the westward shift of the back
454
trajectories, air masses pass through Sandakan at around 16:00 UTC, approximately the time of the sampled spike in V. The
455
shipping activity and oil depots present in this area may be responsible for the spike in oil combustion tracers, indicating the
456
complexity of aerosol transport in the region as small cities like Sandakan may be a source of significant spikes in aerosol.
457
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
21
458
Figure 10. Determination of 18 September event using (a) HYSPLIT back trajectories, (b, c) Google Maps view of the
459
northeastern coast of Borneo, (d) Density of shipping traffic from Sandakan, Sabah (source: MarineTraffic). Red
460
squares indicate the location of the succeeding plot.
461
20-24 Sept: Clean marine period
462
The first half of the cruise showed the lowest concentrations of elements associated with biomass burning K, S, Si, and
463
Al. Back trajectories during this early period originate from the northern part of Borneo and do not penetrate deeply into the
464
MC until late into the cruise (Fig. 2l). During this period, HYSPLIT back trajectories show that air mass pathways shift away
465
from the Borneo coasts towards open sea (Fig. 2j). In addition to the shift away from biomass burning sites, back trajectories
466
between 22 and 24 Sept pass through areas of open sea that host lower levels of shipping traffic (Fig. S2, Supplementary
467
Information).
468
24-26 Sept: Large mixed aerosol event from northwest Borneo
469
Around 26 Sept, increases in fine mode V and Ni occurred when air masses passed through the northwest coast of Borneo,
470
suggesting the presence of ports or oil depots like with the aforementioned spike on 18 Sept from Sandakan. Back trajectories
471
generated every 6 hours starting from 24 Sept 15:00 UTC until 26 Sept 09:00 UTC show little change over this period (not
472
shown) and intersect with the shipping route hub located along northwest Borneo which would explain the V and Ni spikes
473
(Fig. 2k, S1, Supplementary Information). The enrichments of biomass burning and combustion tracers K and S in the sampled
474
air mass span a wider period beginning on 24 Sept until 26 Sept. This may be due to burning activity along the coast of Borneo
475
which hosts several MODIS-detected active fire hotspots. Late-night land breeze from the island may have advected polluted
476
air masses towards the coast.
477
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
22
28-30 Sept: Large biomass burning event from Southern Kalimantan
478
Enhancements of these elements after 28 Sept coincide with a regional increase in AOD (Fig. 2d) and are captured by
479
the NAAPS model in the form of a large smoke event advected northeast (Fig. 2h). Linear regressions show this large aerosol
480
event at the end of the cruise as a distinct group of points with enhanced concentrations of K and S (Fig. S7, Supplementary
481
Information), suggesting an increase in biomass burning activity during this time. Reid et al. (2015) observed a sharp increase
482
in the number of active fire hotspots, particularly in Sumatra and Southern Kalimantan. As discussed prior and depicted in Fig.
483
2, TC Nesat played a major in role in synoptic wind patterns during the cruise, causing a shift in back trajectories after 28 Sept
484
to the southwest coast of Borneo island. Thus, the enhancements of submicron K, S, Si and Al likely originate from biomass
485
burning in the MC.
486
6. Summary and conclusions
487
This study describes the size-resolved aerosol elemental composition of particles collected by a DRUM rotating impactor
488
during the 17 to 30 September 2011 M/Y Vasco cruise in the vicinity of the Palawan island of the Philippines. This region was
489
chosen due to its location as a receptor for MC aerosol sources, such as biomass burning, oil combustion and soil dust.
490
Meteorological conditions during the cruise were conducive to southwesterly long range transport for seasonal burning aerosol
491
which was observed in the concentration time series of tracers and satellite-derived AOD. Size-resolved aerosol composition
492
in the coarse (1.15-10 μm), fine (0.34- 1.15 μm) and ultrafine (0.07-0.34 μm) modes were used as key tracers to ascertain
493
source contributions. Despite the meteorological complexity of the SCS/WPS, we can gain insights into aerosol sources by
494
focusing on key elemental species. The time series of key elements showed distinct events on 18-19 Sept, 24-26 Sept, and 28-
495
30 Sept, with clean aerosol periods between events. These aerosol events served as case studies of sources in the region. While
496
biomass burning is indeed a key source of aerosol, other sources such as oil combustion, sea spray, fly ash, and soil dust
497
contribute to the chemical profile of the SCS/WPS during the southwest monsoon. Understanding these sources is key to
498
characterizing aerosol composition and transport in the SCS/WPS and, by extension, developing our understanding of aerosol-
499
cloud behavior in the region. As back trajectory analysis and aerosol chemistry showed the presence of multiple key sources,
500
the general conclusions of the study show that:
501
1. Mass distributions of key elements showed the evolution of aerosol chemistry throughout the cruise and
502
interesting covariances between modes. Stage 5 (0.56-0.75 μm) and stage 7 (0.26-0.34 μm) showed enhanced
503
peaks in several elements associated with combustion. Throughout the cruise, mass distributions of V and Ni
504
track each other well both temporally and across DRUM stages, indicative of oil combustion. Mass distributions
505
of V and Ni show higher values in the ultrafine mode between 18-19 September, indicative of an early oil
506
combustion-enriched air mass which was identified to possibly originate from Sandakan, Sabah in Borneo. Mass
507
distributions of K, Al and S show large enhancements in the fine and ultrafine modes after 27 September
508
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
23
coincident with a large aerosol event reported by Reid et al. (2015). In combination with the rapid spread of high
509
AOD and NAAPS-modelled smoke concentration across the region, the strong peaks of these elements at the end
510
of the cruise provide evidence for high levels of MC burning at the end of the cruise. Coarse-mode soil dust
511
elements such as Fe and Si showed similarly-timed enhancements, attributed to soil particle entrainment during
512
burning.
513
2. Short-term meteorological events such as TC Nesat played a key role in long-range transport as they propagated
514
through the region, expediting the northeastward advection of aerosol emissions, an effect observed in previous
515
studies (Atwood et al., 2012; Reid et al., 2012, 2015). The sudden variations in aerosol concentration after 24
516
Sept can be connected to the movement of TC Nesat through the region. Prior to these events, aerosol
517
concentrations remained at generally low levels as NAAPS shows smoke was largely constrained to the southern
518
hemisphere. The passage of TC Nesat advected air masses more northward, allowing them to penetrate deep
519
enough into the northern hemisphere to be sampled by the Vasco. The TC’s passage coincided with a shift in air
520
mass origin from mainland Malaysia prior to 24 Sept to areas known for intense burning activity, most notably
521
Southern Kalimantan by the end of the cruise. This corresponded to a mixed aerosol event from 24 to 26 Sept
522
attributed to Brunei, Borneo and a significant increase in biomass burning tracer concentrations from 28 to 30
523
Sept attributed to Southern Kalimantan. Between these aerosol events, a clean marine event from 26 until 28 Sept
524
was characterized by high concentrations of Cl and low levels of elements associated with anthropogenic sources.
525
Back trajectories showed that air masses travelled through the open, central SCS/WPS which suggest nearly pure
526
sea spray was sampled.
527
3. Five sources across the three modes were resolved by the PMF analysis: biomass burning, oil combustion, soil
528
dust, sea spray, and fly ash. A threshold Pearson R coefficient of 0.0 was used to filter species included in the
529
PMF analysis to improve the interpretability of the PMF solution. Results show that natural sources, sea spray
530
and soil dust, were observed in only the coarse and fine modes while anthropogenic sources, biomass burning,
531
oil combustion, and fly ash, were resolved purely in the fine and ultrafine modes. A strong correspondence
532
between key elements seen on the PMF source profiles and groupings of these elements on the correlation
533
matrices adds confidence to the PMF solution. The biomass burning PMF factor showed the highest percent
534
contributions to total elemental PM mass: 34% in the fine mode, and 43% in the ultrafine mode. It is interesting
535
to note that the contribution of the oil combustion factor increased significantly towards finer modes, 9% in the
536
fine mode but 29% in the ultrafine mode, corroborating its anthropogenic identification. In terms of aerosol
537
events, PMF source contributions were able to capture the most events seen in the raw elemental concentrations.
538
Differences in the temporal variations between PMF-resolved sources suggest these sources are distinct.
539
However, PMF did not differentiate between an early ultrafine Si spike from a distinct, subsequent spike in V
540
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
24
which demonstrates that PMF may merge events, leading to a loss in resolution as observed in other studies (Van
541
Pinxteren et al., 2016). This, however, can be ameliorated with an in-depth, supervised analysis of the data as
542
done in this study.
543
4. As stated above, spikes in oil combustion tracers V and Ni were observed on 18 Sept in the fine and ultrafine
544
modes. HYSPLIT back trajectories suggest the origin of the air mass as Sandakan, an industrial area and port
545
city of Sabah known for its oil depots and shipping activity located along the northeastern coast of Borneo. The
546
spike in oil combustion suggest that a small city can cause drastic increases in tracer concentration depending on
547
air mass trajectories. The strong presence of ultrafine mode Si from 18-19 September was also observed but the
548
time series of Si is distinct from the time series of V and Ni, suggestive of a source distinct from oil combustion.
549
5. The 24 to 26 September event coincided with the arrival of TC Nesat east of Luzon (northeast of the Vasco’s
550
location). Enhancements of multiple key tracers for biomass burning, oil combustion and soil dust were observed,
551
indicative of aerosols mixing within an air mass during transport. Biomass burning tracers K, S, Si, Al show
552
enhancements over a wider period (24-26 Sept) than that of oil combustion tracers V and Ni, which spiked at the
553
end of the period. Furthermore, aerosol-convection interactions were observed as sharp dips in the concentrations
554
of biomass burning and soil dust tracers around 25 Sept before recovery. Interestingly, this dip was not observed
555
for oil combustion tracers V, Ni. This cold pool event was reported in detail by Reid et al. (2015) and this study
556
further elaborated on its impact on PM of different elemental composition. This case demonstrates the effect of
557
short-term or high frequency phenomena on aerosol transport in the MC. HYSPLIT back trajectories show that
558
air masses begin to travel from the southwest MC in response to TC Nesat’s inflow arm. Air masses during the
559
24-26 September event pass through Brunei, a shipping hub located along the northeastern coast of Borneo,
560
which explains the increase in oil combustion tracers V and Ni. The coast was also observed to host a number of
561
active fire hotspots. Land breeze may lead to the addition of burning plumes into the traveling air mass which
562
would explain the enrichment.
563
6. The 28-30 September aerosol event showed an enrichment in K and S that coincided with a shift in back trajectory
564
origin to Southern Kalimantan, which hosts a high fire hotspot density. MC burning may be characterized by an
565
elevated K/S ratio and strong fine and ultrafine mode peaks in the mass distributions of S and K. The 28-30
566
September event also coincided with the enhancement of soil dust elements in the coarse mode, indicative of soil
567
particle entrainment during burning activity (Reid et al., 2015).
568
The study identified source locations of aerosol and characterized the plumes during the Vasco 2011 cruise; however,
569
unanswered questions remain such as the origin of the strong ultrafine Si signal detected early in the cruise (18-19 Sept) which
570
may be connected to a rapid local nucleation event. The source location of the PMF-resolved fly ash factor also remains
571
unidentified due to its complicated source contribution time series and unclear elemental profile. Investigation into cloud nuclei
572
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
25
(CN) properties during the cruise may be done to further validate the intensity and timing of plumes. In addition to the findings
573
of this study on the elemental PM, future research on other species collected during the 2011 and 2012 Vasco campaigns such
574
as trace gases may compliment and deepen our current understanding of the aerosol environment in the SCS/WPS by adding
575
more degrees of freedom, specifically the lifetimes of trace gases and potential for secondary aerosol formation during
576
transport.
577
Author contribution
578
MRAH performed the analysis and prepared the manuscript. MTC supervised the analysis, especially for the PMF
579
section. MOLC supervised the analysis and provided input for the manuscript. JSR collected the data onboard the Vasco,
580
supervised the analysis, provided input for the manuscript. PX provided the NAAPS Smoke model outputs for Fig. 2 and
581
provided input for the manuscript. JBS, NDL, SNYU collected the data onboard the Vasco. SC, YJZ performed the XRF
582
analysis on the data.
583
Data availability
584
The Vasco ship data is available through correspondence with Jeffrey S. Reid, jeffrey.reid@nrlmry.navy.mil. MODIS
585
AOD images were obtained from the NASA Worldview application: https://worldview.earthdata.nasa.gov/. HYSPLIT data is
586
accessible through the NOAA READY website (http://www.ready.noaa.gov). NAAPS aerosol reanalysis data can be accessed
587
at the US GODAE server: http://www.usgodae.org/.
588
Competing Interests
589
The authors declare that they have no conflict of interest.
590
Acknowledgements
591
We acknowledge the use of imagery from the NASA Worldview application (https://worldview.earthdata.nasa.gov/),
592
part of the NASA Earth Observing System Data and Information System (EOSDIS). The authors gratefully acknowledge the
593
NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY
594
website (http://www.ready.noaa.gov) used in this publication.
595
References
596
Alam, K., Mukhtar, A., Shahid, I., Blaschke, T., Majid, H., Rahman, S., Khan, R. and Rahman, N.: Source apportionment
597
and characterization of particulate matter (PM10) in urban environment of Lahore, Aerosol Air Qual. Res., 14(7), 1851–
598
1861, doi:10.4209/aaqr.2014.01.0005, 2014.
599
Artaxo, P. and Maenhaut, W.: Aerosol Characteristics and Sources for the Amazon Basin During the Wet Season, J.
600
Geophys. Res., 95(D10), 16971–16985, 1990.
601
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
26
Artaxo, P., Fernandas, E. T., Martins, J. V., Yamasoe, M. A., Maenhaut, W., Longo, K. M., Castanho, A. and Hobbs, P. V.:
602
Large-scale aerosol source apportionment in Amazonia, J. Geophys. Res. Atmos., 103(D24), 31837–31847,
603
doi:10.1029/98jd02346, 1998.
604
Atwood, S. A., Reid, J. S., Kreidenweis, S. M., Cliff, S. S., Zhao, Y., Lin, N. H., Tsay, S. C., Chu, Y. C. and Westphal, D.
605
L.: Size resolved measurements of springtime aerosol particles over the northern South China Sea, Atmos. Environ., 78,
606
134–143, doi:10.1016/j.atmosenv.2012.11.024, 2012.
607
Atwood, S. A., Salinas, S. V., Chew, B. N., Reid, J. S., Balasubramanian, R., Yu, L. E. and Kreidenweis, S. M.: Analysis of
608
source regions for smoke events in Singapore for the 2009 El Nino burning season, Atmos. Environ., 78, 219–230,
609
doi:10.1016/j.atmosenv.2013.04.047, 2013.
610
Atwood, S. A., Reid, J. S., Kreidenweis, S. M., Blake, D. R., Jonsson, H. H., Lagrosas, N. D., Xian, P., Reid, E. A., Sessions,
611
W. R. and Simpas, J. B.: Size-resolved aerosol and cloud condensation nuclei (CCN) properties in the remote marine
612
South China Sea-Part 1: Observations and source classification, Atmos. Chem. Phys., 17(2), 1105–1123,
613
doi:10.5194/acp-17-1105-2017, 2017.
614
Baker, K. R., Simon, H. and Kelly, J. T.: Challenges to modeling “cold pool” meteorology associated with high pollution
615
episodes, Environ. Sci. Technol., 45(17), 7118–7119, doi:10.1021/es202705v, 2011.
616
Balasubramanian, R., Qian, W.-B., Decesari, S., Facchini, M. C. and Fuzzi, S.: Comprehensive characterization of PM 2.5
617
aerosols in Singapore, J. Geophys. Res., 108(D16), doi:10.1029/2002jd002517, 2003.
618
Becagli, S., Sferlazzo, D. M., Pace, G., Di Sarra, A., Bommarito, C., Calzolai, G., Ghedini, C., Lucarelli, F., Meloni, D.,
619
Monteleone, F., Severi, M., Traversi, R. and Udisti, R.: Evidence for heavy fuel oil combustion aerosols from chemical
620
analyses at the island of Lampedusa: A possible large role of ships emissions in the Mediterranean, Atmos. Chem. Phys.,
621
12(7), 3479–3492, doi:10.5194/acp-12-3479-2012, 2012.
622
Boucher, O., D. Randall, P. Artaxo, C. Bretherton, G. Feingold, P. Forster, V.-M. Kerminen, Y. Kondo, H. Liao, U.
623
Lohmann, P. Rasch, S.K. Satheesh, S. Sherwood, B. Stevens and X.Y. Zhang: Clouds and Aerosols. In: Climate Change
624
2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the
625
Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung,
626
A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New
627
York, NY, USA, 2013.
628
Cahill, T. A., Barnes, D. E., Lawton, J. A., Miller, R., Spada, N., C, R. D. W. and Kimbrough, S.: Transition metals in
629
coarse, fine, very fine and ultra-fine particles from an interstate highway transect near Detroit, , doi:10.1016/S0065-
630
3454(07)37008-3, 2007.
631
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
27
Chan, Y. C., Christensen, E., Golding, G., King, G., Gore, W., Cohen, D. D., Stelcer, E., Simpson, R., Denison, L. and
632
Wong, N.: Source Apportionment of Ambient Volatile Organic Compounds in Major Cities in Australia By Positive
633
Matrix Factorisation , 42(2), 22–29, 2008.
634
Chang, C. P., Zhuo, W., John, M. and Ching-Hwang, L.: Annual Cycle of Southeast Asia — Maritime Continent Rainfall
635
and the Asymmetric, J. Clim., 18, 287–301, doi:10.1175/JCLI-3257.1, 2005.
636
Cohen, D. D., Crawford, J., Stelcer, E. and Bac, V. T.: Long range transport of fine particle windblown soils and coal fired
637
power station emissions into Hanoi between 2001 to 2008, Atmos. Environ., 44(31), 3761–3769,
638
doi:10.1016/j.atmosenv.2010.06.047, 2010a.
639
Cohen, D. D., Crawford, J., Stelcer, E. and Bac, V. T.: Characterisation and source apportionment of fine particulate sources
640
at Hanoi from 2001 to 2008, Atmos. Environ., 44(3), 320–328, doi:10.1016/j.atmosenv.2009.10.037, 2010b.
641
Davison, R. L., Natusch, D. F. S. and Wallace, J. R.: Trace Elements in Fly Ash, , 8(13) [online] Available from:
642
https://pubs.acs.org/doi/pdf/10.1021/es60098a003?src=recsys, 1974.
643
Deonarine, A., Kolker, A. and Doughten, M.: Trace Elements in Coal Ash, Usgs, (May), 1–6,
644
doi:10.1002/9781119199090.fmatter, 2015.
645
Draxler, R. R. and Hess, G. D.: An overview of the HYSPLIT_4 modelling system for trajectories, Aust. Meteorol. Mag.,
646
47, 295–308, 1998.
647
Draxler, R. R.: HYSPLIT4 user’s guide. NOAA Tech. Memo. ERL ARL -230, NOAA Air Resources Laboratory, Silver
648
Spring, MD., 1999.
649
Duncan, B. N., Martin, R. V., Staudt, A. C., Yevich, R. and Logan, J. A.: Interannual and seasonal variability of biomass
650
burning emissions constrained by satellite observations, J. Geophys. Res., 108(D2), doi:10.1029/2002jd002378, 2003.
651
Engling, G., He, J., Betha, R. and Balasubramanian, R.: Assessing the regional impact of Indonesian biomass burning
652
emissions based on organic molecular tracers and chemical mass balance modeling, Atmos. Chem. Phys., 14(15), 8043–
653
8054, doi:10.5194/acp-14-8043-2014, 2014.
654
Farao, C., Canepari, S., Perrino, C. and Harrison, R. M.: Sources of PM in an industrial area: Comparison between receptor
655
model results and semiempirical calculations of source contributions, Aerosol Air Qual. Res., 14(6), 1558–1572,
656
doi:10.4209/aaqr.2013.08.0281, 2014.
657
Field, R. D. and Shen, S. S. P.: Predictability of carbon emissions from biomass burning in Indonesia from 1997 to 2006, J.
658
Geophys. Res. Biogeosciences, 113(4), 1–17, doi:10.1029/2008JG000694, 2008.
659
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
28
Fujii, Y., Tohno, S., Amil, N., Latif, M. T., Oda, M., Matsumoto, J. and Mizohata, A.: Annual variations of carbonaceous
660
PM2.5in Malaysia: Influence by Indonesian peatland fires, Atmos. Chem. Phys. Discuss., 15(16), 22419–22449,
661
doi:10.5194/acpd-15-22419-2015, 2015.
662
Ge, C., Wang, J. and Reid, J. S.: Mesoscale modeling of smoke transport over the Southeast Asian maritime continent:
663
Coupling of smoke direct radiative effect below and above the low-level clouds, Atmos. Chem. Phys., 14(1), 159–174,
664
doi:10.5194/acp-14-159-2014, 2014.
665
Gugamsetty, B., Wei, H., Liu, C. N., Awasthi, A., Tsai, C. J., Roam, G. D., Wu, Y. C. and Chen, C. F.: Source
666
Characterization and Apportionment of PM10, PM2.5 and PM0.1 by Using Positive Matrix Factorization, Aerosol Air
667
Qual. Res., 12(4), 476–491, doi:10.4209/aaqr.2012.04.0084, 2012.
668
Han, J. S., Moon, K. J., Lee, S. J., Ryu, S. Y., Cliff, S. S., Kim, Y. J. and Yi, S. M.: Size-resolved source apportionment of
669
ambient particles by positive matrix factorization at Gosan background site in East Asia, Atmos. Chem. Phys., 6(1), 211–
670
223, doi:10.5194/acp-6-211-2006, 2006.
671
Hedberg, E., Gidhagen, L. and Johansson, C.: Source contributions to PM10 and arsenic concentrations in Central Chile
672
using positive matrix factorization, Atmos. Environ., 39(3), 549–561, doi:10.1016/j.atmosenv.2004.11.001, 2005.
673
Higgins, R. W. and Shi, W.: Intercomparison of the principal modes of interannual and intraseasonal variability of the North
674
American Monsoon System, J. Clim., 14(3), 403–417, doi:10.1175/1520-0442(2001)014<0403:IOTPMO>2.0.CO;2,
675
2001.
676
Hoek, G., Krishnan, R. M., Beelen, R., Peters, A., Ostro, B., Brunekreef, B. and Kaufman, J. D.: Long-term air pollution
677
exposure and cardio-respiratory mortality: A review, Environ. Heal. A Glob. Access Sci. Source, 12(1),
678
doi:10.1186/1476-069X-12-43, 2013.
679
Hsu, S. C., Tsai, F., Lin, F. J., Chen, W. N., Shiah, F. K., Huang, J. C., Chan, C. Y., Chen, C. C., Liu, T. H., Chen, H. Y.,
680
Tseng, C. M., Hung, G. W., Huang, C. H., Lin, S. H. and Huang, Y. T.: A super Asian dust storm over the East and South
681
China Seas: Disproportionate dust deposition, J. Geophys. Res. Atmos., 118(13), 7169–7181, doi:10.1002/jgrd.50405,
682
2013.
683
Hyer, E. J., Reid, J. S. and Zhang, J.: An over-land aerosol optical depth data set for data assimilation by filtering, correction,
684
and aggregation of MODIS Collection 5 optical depth retrievals, Atmos. Meas. Tech., 4(3), 379–408, doi:10.5194/amt-4-
685
379-2011, 2011.
686
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
29
Karanasiou, A. A., Siskos, P. A. and Eleftheriadis, K.: Assessment of source apportionment by Positive Matrix Factorization
687
analysis on fine and coarse urban aerosol size fractions, Atmos. Environ., 43(21), 3385–3395,
688
doi:10.1016/j.atmosenv.2009.03.051, 2009.
689
Kecorius, S., Madueno, L., Vallar, E., Alas, H., Betito, G., Birmili, W., Cambaliza, M. O., Catipay, G., Gonzaga-Cayetano,
690
M., Galvez, M. C., Lorenzo, G., Müller, T., Simpas, J. B., Tamayo, E. G. and Wiedensohler, A.: Aerosol particle mixing
691
state, refractory particle number size distributions and emission factors in a polluted urban environment: Case study of
692
Metro Manila, Philippines, Atmos. Environ., 170, 169–183, doi:10.1016/j.atmosenv.2017.09.037, 2017.
693
Landis, M. S., Pancras, J. P., Graney, J. R., White, E. M., Edgerton, E. S., Legge, A. and Percy, K. E.: Source apportionment
694
of ambient fine and coarse particulate matter at the Fort McKay community site, in the Athabasca Oil Sands Region,
695
Alberta, Canada, Sci. Total Environ., 584–585, 105–117, doi:10.1016/j.scitotenv.2017.01.110, 2017.
696
Lareau, N. P., Crosman, E., Whiteman, C. D., Horel, J. D., Hoch, S. W., Brown, W. O. J. and Horst, T. W.: The persistent
697
cold-air pool study, Bull. Am. Meteorol. Soc., 94(1), 51–63, doi:10.1175/BAMS-D-11-00255.1, 2013.
698
Lee, E., Chan, C. K. and Paatero, P.: Application of positive matrix factorization in source apportionment of particulate
699
pollutants in Hong Kong, Atmos. Environ., 33(19), 3201–3212, doi:10.1016/S1352-2310(99)00113-2, 1999.
700
Lee, S.-S., Feingold, G. and Chuang, P. Y.: Effect of Aerosol on Cloud–Environment Interactions in Trade Cumulus, J.
701
Atmos. Sci., 69(12), 3607–3632, doi:10.1175/jas-d-12-026.1, 2012.
702
Lestari, P. and Mauliadi, Y. D.: Source apportionment of particulate matter at urban mixed site in Indonesia using PMF,
703
Atmos. Environ., 43(10), 1760–1770, doi:10.1016/j.atmosenv.2008.12.044, 2009.
704
Li, T., Wang, Y., Li, W. J., Chen, J. M., Wang, T. and Wang, W. X.: Concentrations and solubility of trace elements in fine
705
particles at a mountain site, southern China: Regional sources and cloud processing, Atmos. Chem. Phys., 15(15), 8987–
706
9002, doi:10.5194/acp-15-8987-2015, 2015.
707
Liao, H.-T., Chou, C. C.-K., Chow, J. C., Watson, J. G., Hopke, P. K. and Wu, C.-F.: Source and risk apportionment of
708
selected VOCs and PM2.5 species using partially constrained receptor models with multiple time resolution data,
709
Environ. Pollut., 205, 121–130, doi:10.1016/j.envpol.2015.05.035, 2015.
710
Lin, C.-Y., Wang, Z., Chen, W.-N., Chang, S.-Y., Chou, C. C. K., Sugimoto, N. and Zhao, X.: Long-range transport of Asian
711
dust and air pollutants to Taiwan: observed evidence and model simulation, Atmos. Chem. Phys. Discuss., 6(5), 10183–
712
10216, doi:10.5194/acpd-6-10183-2006, 2007.
713
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
30
Lin, Y.-C., Tsai, C.-J., Wu, Y.-C., Zhang, R., Chi, K.-H., Huang, Y.-T., Lin, S.-H. and Hsu, S.-C.: Characteristics of trace
714
metals in traffic-derived particles in Hsuehshan Tunnel, Taiwan: size distribution, potential source, and fingerprinting
715
metal ratio, Atmos. Chem. Phys., 15(8), 4117–4130, doi:10.5194/acp-15-4117-2015, 2015.
716
Loo, Y. Y., Billa, L. and Singh, A.: Effect of climate change on seasonal monsoon in Asia and its impact on the variability of
717
monsoon rainfall in Southeast Asia, Geosci. Front., 6(6), 817–823, doi:10.1016/j.gsf.2014.02.009, 2015.
718
Lynch, P., Reid, J. S., Westphal, D. L., Zhang, J., Hogan, T. F., Hyer, E. J., Curtis, C. A., Hegg, D. A., Shi, Y., Campbell, J.
719
R., Rubin, J. I., Sessions, W. R., Turk, F. J. and Walker, A. L.: An 11-year global gridded aerosol optical thickness
720
reanalysis (v1.0) for atmospheric and climate sciences, Geosci. Model Dev., 9(4), 1489–1522, doi:10.5194/gmd-9-1489-
721
2016, 2016.
722
Maloney, E. D. and Hartmann, D. L.: The Madden–Julian Oscillation, Barotropic Dynamics, and North Pacific Tropical
723
Cyclone Formation. Part I: Observations, J. Atmos. Sci., 58(17), 2545–2558, doi:10.1175/1520-
724
0469(2001)058<2545:TMJOBD>2.0.CO;2, 2001.
725
Mazzei, F., D’Alessandro, A., Lucarelli, F., Nava, S., Prati, P., Valli, G. and Vecchi, R.: Characterization of particulate
726
matter sources in an urban environment, Sci. Total Environ., 401(1–3), 81–89, doi:10.1016/j.scitotenv.2008.03.008,
727
2008.
728
Markowski, G. R. and Filby, R.: Trace Element Concentration as a Function of Particle Size in Fly Ash from a Pulverized
729
Coal Utility Boiler, Environ. Sci. Technol., 19(9), 796–804, doi:10.1021/es00139a005, 1985.
730
Nakajima, T., Yoon, S. C., Ramanathan, V., Shi, G. Y., Takemura, T., Higurashi, A., Takamura, T., Aoki, K., Sohn, B. J.,
731
Kim, S. W., Tsuruta, H., Sugimoto, N., Shimizu, A., Tanimoto, H., Sawa, Y., Lin, N. H., Lee, C. Te, Goto, D. and
732
Schutgens, N.: Overview of the Atmospheric Brown Cloud East Asian Regional Experiment 2005 and a study of the
733
aerosol direct radiative forcing in east Asia, J. Geophys. Res. Atmos., 112(24), 1–23, doi:10.1029/2007JD009009, 2007.
734
Nigam, A., Welch, W., Wayne Miller, J., and Cocher III, D. R.: Effect of fuel sulphur content and control technology on PM
735
emission from ship’s auxiliary engine. Proceeding international aerosol conference, St. Paul, USA, 10–15 September
736
2006, 1531–1532, 2006.
737
Oanh, Nguyen Thi Kim. Integrated air quality management: Asian case studies. Boca Raton, FL: Taylor & Francis, 2013.
738
Print.
739
O’Sullivan, D.Murray, B. J., Malkin, T. L., Whale, T. F., Umo, N. S., Atkinson, J. D., Price, H. C., Baustian, K. J., Browse,
740
J. and Webb, M. E.: Ice nucleation by fertile soil dusts: relative importance of mineral and biogenic components, Atmos.
741
Chem. Phys., 14(4), 1853–1867, doi:10.5194/acp-14-1853-2014, 2014.
742
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
31
Paatero, P. and Tapper, U.: Positive matrix factorization: A non‐negative factor model with optimal utilization of error
743
estimates of data values, Environmetrics, 5(2), 111–126, doi:10.1002/env.3170050203, 1994.
744
Paatero, P. and Hopke, P. K.: Discarding or downweighting high-noise variables in factor analytic models, Anal. Chim.
745
Acta, 490(1–2), 277–289, doi:10.1016/S0003-2670(02)01643-4, 2003.
746
Paterson, K. G., Sagady, J. L., Hooper, D. L., Bertman, S. B., Carroll, M. A. and Shepson, P. B.: Analysis of air quality data
747
using positive matrix factorization, Environ. Sci. Technol., 33(4), 635–641, doi:10.1021/es980605j, 1999.
748
Polissar, A. V., Hopke, P. K., Paatero, P., Malm, W. C. and Sisler, J. F.: Atmospheric aerosol over Alaska 2. Elemental
749
composition and sources, J. Geophys. Res. Atmos., 103(D15), 19045–19057, doi:10.1029/98JD01212, 1998.
750
Prospero, J. M.: Long-range transport of mineral dust in the global atmosphere: Impact of African dust on the environment
751
of the southeastern United States, Proc. Natl. Acad. Sci., 96(7), 3396–3403, doi:10.1073/pnas.96.7.3396, 1999.
752
Qi, L., Chen, M., Ge, X., Zhang, Y. and Guo, B.: Seasonal variations and sources of 17 aerosol metal elements in suburban
753
Nanjing, China, Atmosphere (Basel)., 7(12), 1–21, doi:10.3390/atmos7120153, 2016.
754
Qian, J.-H., Robertson, A. W. and Moron, V.: Interactions among ENSO, the Monsoon, and Diurnal Cycle in Rainfall
755
Variability over Java, Indonesia, J. Atmos. Sci., 67(11), 3509–3524, doi:10.1175/2010jas3348.1, 2010.
756
Querol, X., Viana, M., Alastuey, A., Amato, F., Moreno, T., Castillo, S., Pey, J., Rosa, J. de la, Campa, A. S. de la, Artinano,
757
B., Salvador, P., Santos, S. G. Dos, Fernandez-Patier, R., Moreno-Grau, S., Negral, L., Minguillon, M. C., Monfort, E.,
758
Gil, J. I., Inza, A., Ortega, L. A., Santamaria, J. M. and Zabalza, J.: Source origin of trace elements in PM from regional
759
background, urban and industrial sites of Spain, Atmos. Environ., 41(34), 7219–7231,
760
doi:10.1016/j.atmosenv.2007.05.022, 2007.
761
Raabe, O. G., Braaten, D. A., Axelbaum, R. L., Teague, S. V and Cahill, T. A.: Calibration studies of the drum impactor, J.
762
Aerosol Sci., 19(2), 183–195, doi:10.1016/0021-8502(88)90222-4, 1988.
763
Reid, J. S., Flocchini, R. G., Cahill, T. A., Ruth, R. S. and Salgado, D. P.: Local meteorological, transport, and source
764
aerosol characteristics of late autumn Owens Lake (dry) dust storms, Atmos. Environ., 28(9), 1699–1706,
765
doi:10.1016/1352-2310(94)90315-8, 1994.
766
Reid, J. S., Hyer, E. J., Prins, E. M., Westphal, D. L., Zhang, J., Wang, J., Christopher, S. A., Curtis, C. A., Schmidt, C. C.,
767
Eleuterio, D. P., Richardson, K. A. and Hoffman, J. P.: Global monitoring and forecasting of biomass-burning smoke:
768
Description of and lessons from the fire Locating and Modeling of Burning Emissions (FLAMBE) program, IEEE J. Sel.
769
Top. Appl. Earth Obs. Remote Sens., 2(3), 144–162, doi:10.1109/JSTARS.2009.2027443, 2009.
770
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
32
Reid, J. S., Xian, P., Hyer, E. J., Flatau, M. K., Ramirez, E. M., Turk, F. J., Sampson, C. R., Zhang, C., Fukada, E. M. and
771
Maloney, E. D.: Multi-scale meteorological conceptual analysis of observed active fire hotspot activity and smoke optical
772
depth in the Maritime Continent, Atmos. Chem. Phys., 12(4), 2117–2147, doi:10.5194/acp-12-2117-2012, 2012.
773
Reid, J., Hyer, E., Johnson, R., Holben, B. N., Yokelson, R., Zhang, J., Campbell, J., Christopher, S., DiGirolamo, L., Giglio,
774
L., Holz, R., Kearney, C., Miettinen, J., Reid, E., Turk, F. J., Wang, J., Xian, P., Zhao, G., Balasubramanian, R., Chew,
775
B. N., Janjai, S., Lagrosas, N., Lestari, P., Lin, N.-H., Mahmud, M., Nguyen, A. X., Norris, B., Oanth, N. T. K., Oo, M.,
776
Salinas, S. V., Welton, E. J. and Liew, S. C.: Observing and understanding the Southeast Asian aerosol system by remote
777
sensing: An initial review and analysis for the Seven Southeast Asian Studies (7SEAS) program, Atmos. Res., 122, 403–
778
468, doi:10.1016/j.atmosres.2012.06.005, 2013.
779
Reid, J. S., Lagrosas, N. D., Jonsson, H. H., Reid, E. A., Sessions, W. R., Simpas, J. B., Uy, S. N., Boyd, T. J., Atwood, S.
780
A., Blake, D. R., Campbell, J. R., Cliff, S. S., Holben, B. N., Holz, R. E., Hyer, E. J., Lynch, P., Meinardi, S., Posselt, D.
781
J., Richardson, K. A., Salinas, S. V., Smirnov, A., Wang, Q., Yu, L. and Zhang, J.: Observations of the temporal
782
variability in aerosol properties and their relationships to meteorology in the summer monsoonal South China Sea/East
783
Sea: the scale-dependent role of monsoonal flows, the Madden–Julian Oscillation, tropical cyclones, squall lines and cold
784
pools, Atmos. Chem. Phys., 15(4), 1745–1768, doi:10.5194/acp-15-1745-2015, 2015.
785
Reid, J. S., Xian, P., Holben, B. N., Hyer, E. J., Reid, E. A., Salinas, S. V., Zhang, J., Campbell, J. R., Chew, B. N., Holz, R.
786
E., Kuciauskas, A. P., Lagrosas, N., Posselt, D. J., Sampson, C. R., Walker, A. L., Welton, E. J. and Zhang, C.: Aerosol
787
meteorology of the Maritime Continent for the 2012 7SEAS southwest monsoon intensive study – Part 1: regional-scale
788
phenomena, Atmos. Chem. Phys., 16(22), 14041–14056, doi:10.5194/acp-16-14041-2016, 2016.
789
Ross, A. D., Holz, R. E., Quinn, G., Reid, J. S., Xian, P., Turk, F. J. and Posselt, D. J.: Exploring the First Aerosol Indirect
790
Effect over the Maritime Continent Using a Ten-Year Collocated MODIS, CALIOP, and Model Dataset, Atmos. Chem.
791
Phys. Discuss., (March), 1–28, doi:10.5194/acp-2018-4, 2018.
792
Rolph, G., Stein, A., and Stunder, B. Real-time Environmental Applications and Display sYstem: READY. Environmental
793
Modelling & Software, 95, 210-228, https://doi.org/10.1016/j.envsoft.2017.06. 2017.
794
Santoso, M., Lestiani, D. D., Mukhtar, R., Hamonangan, E., Syafrul, H., Markwitz, A. and Hopke, P. K.: Preliminary study
795
of the sources of ambient air pollution in Serpong, Indonesia, Atmos. Pollut. Res., 2(2), 190–196,
796
doi:10.5094/apr.2011.024, 2011.
797
Schlosser, J. S., Braun, R. A., Bradley, T., Dadashazar, H., MacDonald, A. B., Aldhaif, A. A., Aghdam, M. A., Mardi, A. H.,
798
Xian, P. and Sorooshian, A.: Analysis of aerosol composition data for western United States wildfires between 2005 and
799
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
33
2015: Dust emissions, chloride depletion, and most enhanced aerosol constituents, J. Geophys. Res. Atmos., 122(16),
800
8951–8966, doi:10.1002/2017JD026547, 2017.
801
Seneviratne, M. C. S., Waduge, V. A., Hadagiripathira, L., Sanjeewani, S., Attanayake, T., Jayaratne, N. and Hopke, P. K.:
802
Characterization and source apportionment of particulate pollution in Colombo, Sri Lanka, Atmos. Pollut. Res., 2(2),
803
207–212, doi:10.5094/apr.2011.026, 2011.
804
Shi, Y., Zhang, J., Reid, J. S., Liu, B., and Hyer, E. J.: Critical evaluation of cloud contamination in the MISR aerosol
805
products using MODIS cloud mask products, Atmos. Meas. Tech., 7, 1791-1801, 2014.
806
Song, Y., Xie, S., Zhang, Y., Zeng, L., Salmon, L. G. and Zheng, M.: Source apportionment of PM2.5 in Beijing using
807
principal component analysis/absolute principal component scores and UNMIX, Sci. Total Environ., 372(1), 278–286,
808
doi:10.1016/j.scitotenv.2006.08.041, 2006.
809
Sorooshian, A., Feingold, G., Lebsock, M. D., Jiang, H., and Stephens, G. L.: On the precipitation susceptibility of clouds to
810
aerosol perturbations, Geophys. Res. Lett., 36, L13803, doi:10.1029/2009GL038993, 2009.
811
Sowlat, M. H., Hasheminassab, S. and Sioutas, C.: Source apportionment of ambient particle number concentrations in
812
central Los Angeles using positive matrix factorization (PMF), Atmos. Chem. Phys., 16(8), 4849–4866, doi:10.5194/acp-
813
16-4849-2016, 2016.
814
Stein, A.F., Draxler, R.R, Rolph, G.D., Stunder, B.J.B., Cohen, M.D., and Ngan, F. NOAA's HYSPLIT atmospheric
815
transport and dispersion modeling system, Bull. Amer. Meteor. Soc., 96, 2059-2077, http://dx.doi.org/10.1175/BAMS-D-
816
14-00110.1, 2015.
817
Thurston, G. D., Ito, K. and Lall, R.: A Source Apportionment of U.S. Fine Particulate Matter Air Pollution, Atmos.
818
Environ., 45(24), 3831–3840, doi: 10.1016/j.atmosenv.2011.04.070, 1994.
819
Tian, S. L., Pan, Y. P. and Wang, Y. S.: Size-resolved source apportionment of particulate matter in urban Beijing during
820
haze and non-haze episodes, Atmos. Chem. Phys., 16(1), 1–19, doi:10.5194/acp-16-1-2016, 2016.
821
Van Pinxteren, D., Fomba, K. W., Spindler, G., Müller, K., Poulain, L., Iinuma, Y., Löschau, G., Hausmann, A. and
822
Herrmann, H.: Regional air quality in Leipzig, Germany: Detailed source apportionment of size-resolved aerosol
823
particles and comparison with the year 2000, Faraday Discuss., 189, 291–315, doi:10.1039/c5fd00228a, 2016.
824
Viana, M., Kuhlbusch, T. A. J., Querol, X., Alastuey, A., Harrison, R. M., Hopke, P. K., Winiwarter, W., Vallius, M., Szidat,
825
S., Prévôt, A. S. H., Hueglin, C., Bloemen, H., Wåhlin, P., Vecchi, R., Miranda, A. I., Kasper-Giebl, A., Maenhaut, W.
826
and Hitzenberger, R.: Source apportionment of particulate matter in Europe: A review of methods and results, J. Aerosol
827
Sci., 39(10), 827–849, doi:10.1016/j.jaerosci.2008.05.007, 2008.
828
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
34
Visser, S., Slowik, J. G., Furger, M., Zotter, P., Bukowiecki, N., Canonaco, F., Flechsig, U., Appel, K., Green, D. C.,
829
Tremper, A. H., Young, D. E., Williams, P. I., Allan, J. D., Coe, H., Williams, L. R., Mohr, C., Xu, L., Ng, N. L.,
830
Nemitz, E., Barlow, J. F., Halios, C. H., Fleming, Z. L., Baltensperger, U. and Prévôt, A. S. H.: Advanced source
831
apportionment of size-resolved trace elements at multiple sites in London during winter, Atmos. Chem. Phys., 15(19),
832
11291–11309, doi:10.5194/acp-15-11291-2015, 2015.
833
Wang, B., Wu, R. and Fu, X.: Pacific–East Asian Teleconnection: How Does ENSO Affect East Asian Climate?, J. Clim.,
834
13(9), 1517–1536, doi:10.1175/1520-0442(2000)013<1517:PEATHD>2.0.CO;2, 2000.
835
Wang, S.-H., Tsay, S.-C., Lin, N.-H., Hsu, N. C., Bell, S. W., Li, C., Ji, Q., Jeong, M.-J., Hansell, R. A., Welton, E. J.,
836
Holben, B. N., Sheu, G.-R., Chu, Y.-C., Chang, S.-C., Liu, J.-J. and Chiang, W.-L.: First detailed observations of long-
837
range transported dust over the northern South China Sea, Atmos. Environ., 45(27), 4804–4808,
838
doi:10.1016/j.atmosenv.2011.04.077, 2011.
839
Wang, Z., Sorooshian, A., Prabhakar, G., Coggon, M. M. and Jonsson, H. H.: Impact of emissions from shipping, land, and
840
the ocean on stratocumulus cloud water elemental composition during the 2011 E-PEACE field campaign, Atmos.
841
Environ., 89, 570–580, doi:10.1016/j.atmosenv.2014.01.020, 2014.
842
Webster, P. J., Magana, V. O., Palmer, T. N., Shukla, J., Tomas, R. A., Yanai, M. and Yasunari, T.: Monsoons : Processes ,
843
predictability , and the prospects for prediction 2 . Description of the Monsoons, Weather, 103(C7), 14451–14510, 1998.
844
Weichenthal, S., Villeneuve, P. J., Burnett, R. T., Donkelaar, A. van, Martin, R. V., Jones, R. R., DellaValle, C. T., Sandler,
845
D. P., Ward, M. H. and Hoppin, J. A.: Long-Term Exposure to Fine Particulate Matter: Association with Nonaccidental
846
and Cardiovascular Mortality in the Agricultural Health Study Cohort, Environ. Health Perspect., 122(6), 609–615,
847
doi:10.1289/ehp.1307277, 2014.
848
Wimolwattanapun, W., Hopke, P. K. and Pongkiatkul, P.: Source apportionment and potential source locations of PM2.5 and
849
PM2.5–10 at residential sites in metropolitan Bangkok, Atmos. Pollut. Res., 2(2), 172–181, doi:10.5094/apr.2011.022,
850
2011.
851
Witt, M., Baker, A. R. and Jickells, T. D.: Atmospheric trace metals over the Atlantic and South Indian Oceans:
852
Investigation of metal concentrations and lead isotope ratios in coastal and remote marine aerosols, Atmos. Environ.,
853
40(28), 5435–5451, doi:10.1016/j.atmosenv.2006.04.041, 2006.
854
Wu, C., Larson, T. V., Wu, S., Williamson, J., Westberg, H. H. and Liu, L.-J. S.: Source apportionment of PM2.5 and
855
selected hazardous air pollutants in Seattle, Sci. Total Environ., 386(1–3), 42–52, doi:10.1016/j.scitotenv.2007.07.042,
856
2007.
857
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
35
Xian, P., Reid, J. S., Atwood, S. A., Johnson, R. S., Hyer, E. J., Westphal, D. L. and Sessions, W.: Smoke aerosol transport
858
patterns over the Maritime Continent, Atmos. Res., 122, 469–485, doi:10.1016/j.atmosres.2012.05.006, 2013.
859
Xie, R. K., Seip, H. M., Liu, L. and Zhang, D. S.: Characterization of individual airborne particles in Taiyuan City, China,
860
Air Qual. Atmos. Heal., 2(3), 123–131, doi:10.1007/s11869-009-0039-x, 2009.
861
Xing, Y. F., Xu, Y. H., Shi, M. H. and Lian, Y. X.: The impact of PM2.5 on the human respiratory system, J. Thorac. Dis.,
862
8(1), E69–E74, doi:10.3978/j.issn.2072-1439.2016.01.19, 2016.
863
Yuan, B., Shao, M., Gouw, J. de, Parrish, D. D., Lu, S., Wang, M., Zeng, L., Zhang, Q., Song, Y., Zhang, J. and Hu, M.:
864
Volatile organic compounds (VOCs) in urban air: How chemistry affects the interpretation of positive matrix
865
factorization (PMF) analysis, J. Geophys. Res. Atmos., 117(D24), 1-17, doi:10.1029/2012jd018236, 2012.
866
Yusuf, A. A. and Francisco, H.: Climate change vulnerability mapping for Southeast Asia, Economy and Environment
867
Program for Southeast Asia (EEPSEA) report, available at: http://www. eepsea.net (last access: July 2014), 32 pp., 2009.
868
Zhang, C.: Madden-Julian Oscillation, Rev. Geophys., 43, RG2003, doi:10.1029/2004RG000158, 2005.
869
Zhang, J. and Reid, J. S.: MODIS Aerosol Product Analysis for Data Assimilation: Assessment of Level 2 Aerosol Optical
870
Thickness Retrievals, J. Geophys. Res.-Atmos., 111, 22207, doi:10.1029/2005JD006898, 2006.
871
Zheng, C., Zhao, C., Zhu, Y., Wang, Y., Shi, X., Wu, X., Chen, T., Wu, F. and Qiu, Y.: Analysis of influential factors for the
872
relationship between PM2.5 and AOD in Beijing, Atmos. Chem. Phys., 17(21), 13473–13489, doi:10.5194/acp-17-
873
13473-2017, 2017.
874
Zivanovic, V.: XRF analysis of mineralogical matrix effects and differences between pulverized and fused ferromanganese
875
slag, Chem. Ind. Chem. Eng. Q., 17(2), 231–237, doi:10.2298/ciceq101104008z, 2011.
876
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
36
Table 1. PM1.15/PM10 ratio slopes for elements ordered by ratio-slope.
877
Ratio slope
R-squared
correlation
Ratio average
Standard deviation
V
0.94
0.99
0.95
0.07
K
0.82
0.94
0.35
0.21
S
0.8
0.92
0.49
0.17
Zn
0.74
0.94
0.62
0.04
Y
0.7
0.7
0.53
0.11
Zr
0.7
0.63
0.65
0.07
Mo
0.7
0.67
0.65
0.04
Ti
0.68
0.7
0.53
0.08
Rb
0.61
0.64
0.73
0.09
Al
0.51
0.68
0.55
0.12
Pb
0.47
0.44
0.67
0.06
Cu
0.4
0.42
0.63
0.05
Ni
0.31
0.33
0.61
0.08
As
0.31
0.36
0.33
0.26
Mn
0.3
0.62
0.49
0.19
Si
0.29
0.56
0.32
0.13
Se
0.2
0.24
0.59
0.06
P
0.19
0.32
0.27
0.08
Na
0.16
0.57
0.17
0.03
Sr
0.16
0.11
0.49
0.08
Br
0.13
0.17
0.47
0.08
Ca
0.07
0.59
0.1
0.05
Cl
0.06
0.67
0.04
0.02
Fe
0.06
0.38
0.24
0.12
Mg
0.03
0.29
0.07
0.03
Co
0.03
0.03
0.57
0.1
Ga
0.03
0.04
0.56
0.09
Cr
0.01
0.02
0.19
0.19
878
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
37
Table 2. Sources identified in each size range with PMF. Coarse (1.15-10 μm), fine (0.34-1.15 μm) and ultrafine (0.07-
879
0.34 μm).
880
Source
Major Components
Coarse
Fine
Ultrafine
Biomass Burning
K, S, Si, Al, As
Oil Combustion
V
Sea Spray
Cl, Mg, P, Br
Soil Dust
Fe, Al, Si, Ca, Ti, Zn
Fly ash
As, Se, Pb, Zn, Ti
881
882
https://doi.org/10.5194/acp-2019-352
Preprint. Discussion started: 24 May 2019
c
Author(s) 2019. CC BY 4.0 License.
