Introduction
Given its hypothesized sensitivity to global climate change (e.g., IPCC,
2007; Yusuf and Francisco, 2009), Southeast Asia (SEA) has experienced a substantial increase in
scientific interest; from the region's highly complex meteorology, to its
atmospheric chemistry, air quality, and climate. The region, including the
maritime continent, South China Sea/East Sea (SCS/ES), and Sulu Sea, is
thought to be highly susceptible to aerosol cloud interactions
(Rosenfeld, 1999; Hamid et al., 2001; Yuan et al., 2011). Indeed, around the second half of the boreal summer
monsoonal period from August to mid-October, the seasonal dry climate allows
biomass burning throughout the maritime continent (MC), particularly in warm
El Niño–Southern Oscillation phases (e.g, Nichol, 1998; van der Werf et al., 2004; Field and Shen, 2008; Langner and Siegert, 2009; Field et al., 2009;
van der Kaars et al., 2010; Reid et al., 2012, 2013). Climatologically, there exists both anecdotal evidence
and some station data suggesting an increase in the number of no-rain days
in the Philippines (Cruz et al., 2013), yet perhaps an increase in intense events
(Cinco et al., 2014). Perhaps such a behavior is a result of the effect of increasing
aerosol emissions on clouds. At the same time there is a long-standing
hypothesis that there are increases in mid-level cloudiness, also perhaps
due to increased levels of aerosol particles (Parungo et al., 1994).
Under most circumstances, smoke and pollution from the MC is thought to be
transported by southwesterly monsoonal winds into the SCS/ES where it is
scavenged by convection with eventual annihilation in the monsoonal trough
(Reid et al., 2012; Xian et al., 2013). However, the transition process between “polluted land”
and “clean monsoonal trough” is poorly understood. Large scale modeling
studies suggesting smooth transport are at odds with visible imagery (Reid
et al., 2013) and lidar observations (e.g., Campbell et al., 2013), which suggest smoke is
often sequestered on or very near the major land masses. Owing to near
ubiquitous high cloud cover in the SCS/ES, there are relatively few
satellite observations of smoke transport in the region, except during
anomalously clear or severe events. The limited remote sensing data that are
available is largely qualitative, with both cloud and aerosol retrievals
showing great regional diversity across product lines in this
near-ubiquitous cloud environment (Reid et al., 2013). While
higher-frequency meteorological phenomena, such as the Madden–Julian
Oscillation and equatorial waves (Reid et al., 2012), as well as orographic and sea
breeze effects, are thought to exert significant influence on transport
(Mahmud, 2009a, b; Reid et al., 2012; Wang et al., 2013; Xian et al., 2013), there are virtually no in situ observations
of the SCS/ES aerosol environment in this critical summer monsoonal season.
Cloud processes in regions such as the MC are expected to be sensitive to
the presence of aerosol particles (e.g., Sorooshion et al., 2009; Yuan et al., 2011; Lee et al., 2012). But, we have
little information on how well models perform.
As part of the Seven Southeast Asian Studies (7-SEAS) program (Reid et al., 2013), a 2-week
research cruise was conducted from 17 to 30 September 2011 in the northern
half of the Palawan archipelago of the Philippines; a region thought to be a
long-range receptor for MC biomass burning and industrial emissions (Reid et al., 2012;
Xian et al., 2013). At the same time, additional sun photometer, lidar and ground
measurements were made in Singapore to contrast with the Philippine
receptor. Other sun photometers were located across Southeast Asia.
Conducted on the M/Y Vasco, a locally owned 35 m vessel, our goals were to make
first-ever (to our knowledge) measurements of near-surface aerosol
properties in the region, test the transport hypotheses put forth in Reid et al. (2012), and develop new hypotheses on aerosol–weather interaction that
regulate aerosol prevalence to be studied in future deployments. Most importantly, we aim to develop a narrative on how model simulations and remote sensing retrievals correspond with real world observations in this highly complex aerosol and meteorological environment. Often, the
intricacies of aerosol–meteorological relationships are blurred in bulk
analyses to the detriment of understanding regional physics and chemistry.
Only through studies, such as the one presented here, can we hope to derive the true
sensitivity of the region to aerosol emissions.
In this paper, we give a brief overview of the cruise and its measurements,
as well as other regional measurements made to aid in interpreting the
regional aerosol environment. This will form a descriptive basis for
subsequent 7-SEAS papers on aerosol and cloud features for the 2011 burning
season, as well as a contrast to a similar 2012 cruise to be reported at a
later date. The analysis portion of this paper is focused on the temporal
variability of aerosol particle number and mass concentrations and how these
relate to regional meteorological phenomenon, such as large scale monsoonal
flow, the Madden–Julian Oscillation (MJO), tropical cyclone (TC) development and propagation, and large scale squall
lines/cold pools. We end with a discussion of the strong covariance between
aerosol prevalence and regional thermodynamic behavior, noting how it must
be considered in studies of aerosol, cloud, and precipitation interaction.
Cruise description and instrumentation
This research cruise was conducted on the 35 m, 186 ton M/Y Vasco, owned and
operated by Cosmix Underwater Research Ltd. Manila, Philippines. Photos of
the vessel along with its cruise track are provided in Fig. 1. The Vasco
departed on 17 September 2011 from Navotas, Manila Bay, and returned midday
30 September. The target area for the bulk of the monitoring was in the vicinity
of El Nido and outside of Malampaya Sound, Palawan Island (Lat = 111.1∘ N;
Long = 119.3∘ E). The general mode of operation was to travel to selected
areas, then choose locations for sampling which had a clear breeze to the
open ocean, though protected from the sometimes large swell with no local
wave breaking. Great care was taken to not position the ship downwind of any
sources. Indeed, small settlements are ubiquitous on small islands. But
these were all avoided. The ship would move every 1–2 days within
each area to support other physical oceanographic measurements. The route
south from Manila included a 1-day stop at Apo Reef on 18 September, and the
coast of Culion on 19 September. From 20 September through the morning of
28 September the Vasco operated in the northern Palawan area. On the
morning of 29 September, the Vasco departed El Nido for return to Manila on
the early afternoon of 30 September.
(a) The M/Y Vasco; (b) bow flux tower during the cruise. (c) Map
of cruise area, stars mark key areas of sampling. (d) Enlargement of the
northern Palawan/Coron sampling sites.
Instrumentation was generally deployed in two configuration groups.
Self-contained instrumentation, including meteorology and aerosol chemistry,
was located on a 3 m flux tower on the bow of the ship, a total
top-to-bottom height of 6 m above the ocean surface. This ensured no
self-contamination from the ship except for very rare periods of a following
wind. Aerosol particle counters and nephelometers were located in a forward
locker fed by a 4 cm diameter/4 m long inlet from the top of the ship. Wind
directional data ensured that only periods with air moving over the bow were used
(to remove periods of contamination and self-sampling from the data set).
Periods of self-sampling were also abundantly clear from CN (Condensation Nuclei) counts. Such
periods were obvious, with rapid particle count fluctuations in the 1000 to
10 000+.
Meteorology
The meteorological instrumentation set was associated with the 3 m flux
tower. While fluxes are a subject of a separate paper, a brief summary is
appropriate here. A Campbell sonic anemometer and LI-COR IR H2O/CO2
system were sampled at 50 hz to provide fluxes of momentum, sensible and
latent heat. Mean meteorology was also provided by an R.M. Young propeller
anemometer and a Campbell pressure and ventilated temperature and humidity
probe. Sea surface temperature was provided by a waterline floating
thermocouple. Downwelling shortwave radiation was measured with a Kipp and
Zonen CMP 22 radiometer. Ship location and attitude were given by a Garmin
GPS and accelerometer package. This attitude and velocity data was used to
correct meteorology and solar radiation data.
In addition to the flux tower, ceiling and visibility were provided by a
Vaisala C31 ceilometer, which has been shown to provide information on
aerosol particle profiles when properly corrected (e.g., Clarke et al., 2003; Markowicz et al., 2008;
Tsaknakis et al., 2011). Twenty-five InterMet 1-AB radiosondes were also released during the
cruise, generally one to two per day; 20 of these passed our quality
control. Forward-looking automatic cameras logged images every minute.
Aerosol and gas chemistry
A series of aerosol samplers were mounted on the bow of the ship. One of the
primary instruments utilized in this paper was a free-standing eight-stage
Davis Rotating-drum Uniform size-cut Monitor (DRUM) sampler. The instrument
used in this study was a version of the DRUM sampler originally described by
Cahill et al. (1985), modified to utilize slit orifices and configured to run at 16 L min-1 as described in Reid et al. (2008). A similar instrument was deployed for
comparison to Dongsha Island in the SCS/ES in 2011 in the winter/spring
northeasterly monsoon (Atwood et al., 2013a). An unheated PM10 sample inlet was used
upstream of the impactor, followed by collection stages with nominal 50 %
aerodynamic diameter-cut sizes of 5, 2.5, 1.15, 0.75, 0.56, 0.34, 0.26, and 0.07 µm. Aerosol particles were collected on Mylar strips coated with
Apiezon grease and wrapped around each rotating drum. The drums were rotated
at a consistent rate such that nominal timestamps could be assigned to
specific locations along the strip during compositional analyses, yielding
90 min time resolution. DRUM samples were subjected to X-ray fluorescence
(XRF) analysis at the Advanced Light Source (ALS) of Lawrence Berkeley
National Laboratory to provide measurements of selected elements having
atomic weights between Mg and Mo, along with Pb. Unlike previous DRUM
analyses described in the literature, the XRF analysis samples for this
study utilized a more advanced detector system, making XRF derivations of
key sea salt elements, such as Na and Cl much more quantitative. For
simplicity here, time series of elemental concentration data for the eight
raw-size fractions were combined into two lumped-size fractions: coarse
(stages 1–3 or 10–1.15 µm in aerodynamic size), and fine (stages 4–8,
or 1.15–0.07 µm), respectively. A more detailed analysis will be
provided by a forthcoming paper.
PM2.5 filters were also collected in daily 5 L min-1 MiniVol Tactical Air
Samplers (TASs) and analyzed by gravimetric, XRF and ion chromatography at
the Desert Research Institute. A second set of filters provided organic and
black carbon by the method of Chow et al. (1993). Finally, PM10 and
2.5 samples were collected by the Manila Observatory using both TASs and a
three-stage Dylec impactor for gravimetric and ion chromatography analysis.
These, too, are discussed in Lagrosas et al. (2015).
For trace-gas analysis, 46 whole-air gas samples were collected in
electro-polished stainless steel cans for analysis by gas chromatography by
the University of California Irvine. See Colman et al. (2001) for details, a
full list of 60+ compounds, and relative uncertainties. However, only a
few species are presented here (e.g., CO, and few halo and hydrocarbons).
Flame ionization detectors (FIDs) were used to measure C2–C10
hydrocarbons, electron capture detectors (ECDs) were used for
C1–C2 halocarbons and C1–C5 alkyl nitrates, and
quadrupole mass spectrometer detectors (MSDs) were used for unambiguous
compound identification and selected ion monitoring. Cans were supplied for
the cruise under vacuum, and upon valve release at the ship's bow under
headwind, each collected its volume over the course of ∼ 20 s. Measurement precision varied by species, but was better than 5 %
for the vast majority of species. The most uncertain was
dibromochloromethane at 8 %. Cans were opened sporadically throughout the
cruise, with at least two samples a day being collected, generally in the
morning and afternoon. Sampling was generally not performed during rain
showers. Additional cans were sampled during excellent or interesting
sampling conditions, with the highest frequency during the last few days
when the ship was a receptor for smoke. Of the 46 can samples, 5
did not pass quality assurance as they had anomalously high hydrocarbon and
solvent levels. Given the collection procedure, based over the side on the
windward bow of the ship, we are not entirely sure how the contamination may
have happened, but suspect it may reflect some local contaminant from the
scattered islands in the region. For the purposes of this paper on large
scale flow, they are excluded here.
Ship aerosol microphysics and optics
Onboard the Vasco were a particle counter, sizers, and a nephelometer. Total
particle concentrations were measured by a TSI water condensation nuclei
counter (CPC, http://www.tsi.com/water-based-condensation-particle-counter-3785/). Fine- and coarse-mode particle size was provided by a DMT
(Droplet Measurement Technologies) bench-top Passive Cavity Aerosol Sizing Spectrometer (PCASP), and a TSI
Aerodynamic Particle Sizer (APS) which were calibrated before and after the
cruise. These low-flow rate instruments were behind a Dry-Rite drying
column, which dropped relative humidity to ∼ 50 %. However,
while the CPC and APS operated without incident, the PCASP suffered a relay
failure after the first night at sea (night of 17 September). This was repaired
by 24 September for the second half of the cruise.
For light scattering, we used a TSI three-wavelength nephelometer (λ= 445, 550, 700 nm) at ambient RH, and corrected for
truncation/non-Lambertian light source errors using Anderson et al. (1996).
A three-wavelength particle soot absorption photometer (PSAP) sampled from
the nephelometer stream, and was corrected via Bond et al. (1999). A Radiance Research
single-wavelength nephelometer (λ= 532) was also placed downstream
of the drying column. Finally, a Microtops hand-held sun photometer was
brought on board as part of the Maritime Aerosol Network (MAN; Smirnov et
al., 2011) for measuring aerosol optical thickness (AOT). However, cloudy
skies prohibited measurements prior to the last two days of the cruise (29 and 30 September). Comprehensive studies of aerosol optical properties and size are
a subject of a subsequent paper. However, here we use the CPC and PCASP to
show the time series of basic fine-mode particle number and size properties.
Regional AERONET measurements
In addition to the Vasco cruise, a number of other instruments were placed in the
region to help monitor the aerosol environment. Most notable, in reference
to this paper, was a set of four AERONET sun photometers (Holben et al., 1998), located on
the map in Fig. 2b. Two sites including the Singapore 7-SEAS super site
(e.g., Atwood et al., 2013b), Kuching, Sarawat Borneo (Salinas et al., 2013) and Marbel University,
Mindanao, Philippines were set up for 7-SEAS. Songkhla, Thailand was
pre-existing operational. For the purposes of this paper, we focus one
parameter, 500 nm daily averaged fine-mode AOT. This was generated from the
Level 2.0 spectral deconvolution algorithm (SDA) version 4.1, used to
separate fine- and coarse-mode contributions to AOT (O'Neill et al., 2003). By using the
SDA, we can effectively remove thin cirrus contamination (Chew et al., 2011) and focus
on fine-mode particles from industrial and biomass burning sources.
Overview of the aerosol and meteorological environment during the
17–30 September Vasco cruise. (a) Surface (black) and 850 hPa (purple) NOGAPS
winds overlaid on CMORPH average precipitation rain rates. (b) MODIS
Terra+Aqua active fire hotspot detections during the cruise. Overlaid in
green stars are key AERONET locations. Red star depicts the El Nido receptor
site sampled by the Vasco. (c) Composite average MODIS+MISR aerosol optical
thickness (AOT).
Ancillary satellite and model data
Baseline meteorology data are provided by the Navy Global Atmospheric
Prediction System (NOGAPS; Hogan and Rosmond, 1991). We compared NOGAPS fields to NCAR
reanalysis fields (Kalnay et al., 1996) for the individual events discussed in this paper
and, as we found no substantive differences. NOGAPS data are subsequently
used for initializing the offline Navy Aerosol Analysis and Prediction
System (NAAPS). NAAPS, the Navy's operational aerosol model, is a global
operational 1∘ × 1∘ aerosol transport model supporting
various operations and research, including the monitoring of biomass burning
plumes (Reid et al., 2009). NAAPS has been extensively exercised for the maritime
continent region (e.g., Hyer and Chew, 2010; Reid et al., 2012; Xian et al, 2013). The emissions, transport, and
sinks of a combined pollution product (particulate organic matter plus
sulfates), open biomass burning smoke, and dust are simulated, and
quality-assured AOT retrievals from MODIS observations are assimilated into
the model (Zhang et al., 2008). Model output includes predicted speciated mass
concentrations and AOT. The NAAPS data were used to provide a regional
assessment, as well as along the ship track.
To establish mid- and upper-troposphere air-mass source regions, and the
large scale flow pattern for selected periods of the cruise, back
trajectories were generated using the NOAA Hybrid Single Particle Lagrangian
Integrated Trajectory (HYSPLIT) Version 4.9 Model (Draxler and Hess, 1997, 1998;
Draxler, 2004). The GDAS1, 1∘ × 1∘ global
meteorological data set, generated for HYSPLIT from the Global Data
Assimilation System model, was used to run 72 h backwards trajectories.
Numerous satellite products (visible, IR, cloud heights, scatterometer,
etc.) are also used in an imaging capacity to aid in the analyses. These can
all be found on the NEXSAT system (Miller et al., 2006; http://www.nrlmry.navy.mil/nexsat-bin/nexsat.cgi) and are cited as used in
this paper. We also use other retrieved products for context, such as the
Climate Prediction Center (CPC) MORPHing technique (CMORPH, Joyce et al., 2004) for
precipitation and derived data assimilation-grade satellite AOT products
from MODIS (Zhang et al., 2008) and MISR (Kahn et al., 2009). MODIS fire counts are also used here,
following the regional interpretation of Hyer et al. (2013). While it would
have been highly valuable, Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) data were not collected during the cruise
period due to solar anomalies. However, we do present a single collection from
1 October in conditions we believe to be representative of the last few days of
the cruise.
Results I: regional meteorological and aerosol characteristics
The Vasco cruise occurred in the second half of the month of September 2011.
This period is typically towards the end of the boreal summer southwest
monsoon (henceforth SWM) system, approximately 2–3 weeks before the
transition period to the boreal winter/spring northeast monsoon (NEM). A general
overview of the summer monsoonal system can be found in Chang et al. (2005), Moron et al. (2009) and
the book by Chang et al. (2011). An overview of how monsoonal weather features relate to
smoke emissions and transport from progressively larger to finer scales can
be found in Reid et al. (2012), Xian et al. (2013), Mahmud (2009a, b) and Wang et al. (2013), respectively. A brief
description of key meteorological and aerosol elements for the summer 2011
burning season, as they relate to the study measurement period, is provided
here.
Overall nature of the meteorological and aerosol environment
As discussed in the references above, the SWM in the greater Southeast Asian
region is generally between mid-April and mid-October. Associated
lower-atmospheric flows in the MC are easterly when south of 3∘ S, and
westerly when north of this latitude. In the SCS/ES, surface winds turn
southwesterly, eventually terminating in a monsoonal trough east of the
Philippines. In the upper free troposphere over the SCS/ES, winds flow in
the opposite direction to the marine boundary layer and lower free
troposphere: generally northeasterly, originating from the monsoonal
trough. The ∼ 500 hPa level generally is the delineation
between southwest winds below and northeast winds above. Winds at these
mid-levels are generally light.
For the purposes of this paper, the general meteorology during the cruise is
depicted in Fig. 2a, where NOGAPS surface and 850 hPa winds (black
and magenta, respectively) are provided. These two levels bound the vast
majority of aerosol particles in the region during the SWM (Tosca et al., 2011; Campbell et al., 2013;
Chew et al., 2013; Wang et al., 2013). Average study period precipitation from CMORPH is also
provided as the color background. The red star in the northern Palawan area
indicates the Vasco's position during the bulk of the sampling. Figure 2b
provides a map of all MODIS (Terra+Aqua) fire counts during the study
period. Here, green stars indicate relevant AERONET sun photometer data
utilized in this study. Finally, Fig. 2c provides the average MODIS +
MISR AOTs for the mission, although readers should be aware that AOTs in the
northern half of the domain were derived from only the last few days of the
study when skies were clear enough to perform a retrieval (this is discussed
in more detail later).
The wind fields in the SCS/ES during the study period were largely typical
for the SWM season, with its prevailing southwesterly winds, averaging
∼ 8–20 m s-1 over most of the region. The transition from
easterlies and southeasterlies south of the Equator to southwesterlies in
the SCS/ES can be seen in the general wrapping of the winds around Borneo
and Sumatra. Wind strength anomalies were generally low over the region,
although in the middle of the SCS/ES positive anomalies were on the order of
7 m s-1. Clear cyclonic activity in the northern SCS/ES region is also
apparent. As we discuss later, these positive wind anomalies are result of
TC activity and inflow arm wind enhancement during the cruise. Also notable
is the slight veering wind shear at lowest levels. While the surface winds
are clearly southwesterly, they do become more westerly through the lower
free troposphere to 700 hPa. As discussed later, this has significant
implications for regional aerosol transport and convection.
Precipitation is a maximum along the monsoonal trough, which extends from
the northern SCS/ES to the southeast. However, during the mission,
precipitation was not continuous in this region, but was rather a composite
of enhanced local precipitation, lows, squall lines and tropical cyclone
development. Secondary precipitation maxima were visible and include (1)
convection over land; (2) precipitation west of Sumatra in the so called West
Sumatran low, and (3) convection east of Myanmar driven by convergence of
oceanic air masses reaching land. A depiction of the diversity of regional
cloud features during the mission can be seen in Fig. 3. An area of near
absence of precipitation south of southern Borneo and southern Sumatra
except for isolated mountain top convection, encompassing such islands as
Java and Timor, is a common feature of the SWM.
The 2011 season corresponded to a moderate La Niña year (Multivariate ENSO
(El Niño–Southern Oscillation) Index =-0.95). This typically implies higher precipitation and less fire
activity than normal (Field and Shen, 2008; Field et al., 2009; Reid et al., 2012). However, in this particular year,
precipitation and fire activity were more characteristic of a neutral year.
Thus, while fire activity and smoke AOTs were not akin to the boreal summer
El Niño events of 1997, 2004, 2006, and 2009, the year 2011 ranks in the middle third
in our estimate of fire activity since 2000 (based on Reid et al., 2012, statistics). As is
typical for the late SWM, fire activity was concentrated in southern Sumatra
and southern Borneo/Kalimantan. Fires in this region are often associated
with peatland burning, although a great deal of plantation and small holder
slash burning is common (see Reid et al., 2013 for a discussion of regional burning
practices). As actual peat burning is much more common in drought years
(e.g., Field and Shen, 2008; Miettinen et al., 2010, 2011), we suspect much of the observed burning was
associated with agricultural maintenance or deforestation.
(a) Daily NOGAPS surface winds with CMORPH precipitation for 6
days throughout the cruise demonstrating key meteorological and aerosol
modes. (b) Corresponding NexSat 03:30 UTC/11:30 LST MTSAT visible imagery with
synthetic color background. Ship location at satellite imagery time is
located by a red star.
Intermediate fire activity corresponded with moderate AOT in the region, as
can be seen in Fig. 2c that provides average composites of MISR and MODIS
(Terra+Aqua) AOT. Near the biomass burning sources, AOTs can be high,
averaging over 1 for λ= 550 nm. This is likely low-biased, as AOT
retrievals often flag thick aerosol plumes as cloud in the region (Reid et al., 2013).
Comparison of the Fig. 2 panels elucidates regional transport patterns:
smoke generated in Sumatra and Borneo is carried by the southwesterly winds
through the SCS/ES and eventually scavenged out. Some Sumatran smoke also
crosses the island's western mountain range and enters the Indian Ocean.
While model representation of regional smoke transport often suggests a
smooth transition, imagery – and both passive satellite and lidar
observations – often depict a strong gradient between island and ocean (e.g.,
Campbell et al., 2013; Reid et al., 2012, 2013). Prevailing hypotheses for this divergence surround
scale-dependent issues in the model, and the reproducibility of orographic
and sea breeze meteorology (e.g., Reid et al., 2012; Wang et al., 2013; Xian et al., 2013). But overall, the
transport and transformation mechanisms from polluted island to clean marine
background air are not well understood nor easily simulated. This paper, as
well as subsequent efforts based on this cruise, hope to address these
problems.
Evolution of the meteorological environment during the Vasco cruise
The timing of the Vasco cruise was serendipitous, as it coincided with the
transition of the MJO from wetter to a drier phase in the MC. The MJO is a
large-scale, coupled pattern of meso-synoptic scale circulation and deep
convection that forms in the Indian Ocean and propagates eastward at
∼ 5 m s-1 through and around the MC and into the Pacific
Ocean (Madden and Julian, 1971; Zhang, 2005, 2014). Phase and amplitude of the MJO are quantified
for this study using the method of Wheeler and Hendson (2004). Once this convective region
passes into the central/eastern Pacific and decays, a new event may start in
the Indian Ocean, repeating the cycle. From an aerosol point of view, while
ENSO is an excellent large-scale indicator of seasonal burning, the wet and
dry phases of the MJO strongly influence the intraseasonal timing of
significant smoke events in the MC (Reid et al., 2012). While the MJO was hypothesized
to influence overall AOT (Tian et al., 2008), no satellite-based AOT verification of this
has yet been established due to the difficulty in performing aerosol remote
sensing in the region (Reid et al., 2013). However, fire observations are strongly
enhanced in dry phases (Reid et al., 2012) and mechanistically a
relationship between dry MJO phase, fire emissions, and high AOT seems
certain.
An important correlation of MJO-related convection as it transits and
departs the MC is an associated increase in the formation of regional TCs
(Maloney and Hartman, 2001). Reid et al. (2012) noted that when TCs transit the SCS/ES there is an
increase in both fire activity in the southern MC and ventilation of smoke
into the SCS/ES region. This relationship is thought to be associated with
an acceleration of southwesterly winds in the SCS/ES as air approaches the
TC. As TCs enter the area, strong convection develops along the inflow arm,
scavenging smoke transported offshore. Later, as the TC passes, large-scale
subsidence follows, resulting in negative precipitation anomalies over much
of the SCS/ES and the MC. An example of such a case is presented in global and
mesoscale simulations in Reid et al. (2012) and Wang et al. (2013), respectively. Over the period
of 17–30 September, the MJO convective active phase migrated out of the MC (that
is migrated from Phase 3 to Phase 6) at a relative strength that increased
above the 1 standard deviation intensity level halfway through the period.
The migration of the MJO coincided with a train of TC activity beginning
23 September.
Select examples of daily mean winds with precipitation and representative
daytime MTSAT visible images are found in Fig. 3a, b, respectively.
On 17 September, the day of departure, the general meteorology of the SCS/ES and
the MC was fairly typical for a convectively active phase of the MJO. Regional
lower-tropospheric winds exhibited small anomalies against the NCEP (National
Centers for Environmental Protection) climatology. Comparison of the CMORPH-derived precipitation (Fig. 3a)
with MTSAT visible images (Fig. 3b) suggested the whole region was
showery, with light, scattered precipitation from many small to medium-sized
cells and a few deep and intense storms. Some organization can be seen,
however, in an 800 km wide area in the SCS/ES between southern Vietnam and
Borneo. Over the next 48 h (19 September), precipitation over the
region increased, and the patch of convection in the SCS further organized
and intensified. By 22 September, convection intensified further over the whole
SCS/ES, and cyclonic rotation became clearly evident around a tropical
depression in the northern SCS. This coupled system resulted in lines of
convection and heavier precipitation from the southwest to the northeastern
side of the SCS/ES. The tropical depression was later named Tropical Storm
21 W-Haitang. Haitang continued developing until 25 September, reaching
maximum winds of 18 m s-1. The inflow arm of Haitang moved westward,
leaving the southern SCS/ES drier.
As Haitang was beginning to develop, a separate system, 20 W Nesat, rapidly
intensified in the western Pacific Ocean and migrated westward. As Haitang
then migrated into northern Vietnam, Nesat developed, making landfall on
Luzon on 26 September with maximum 1 min sustained wind speeds of
∼ 58 m s-1 – ultimately listed as a Category 4 TC. After
passing Luzon and causing an estimated USD 1 billion damage, Nesat lost strength to
Category 1 before making landfall again at Hunan Island on 29 September.
Finally, the third tropical cyclone, the westward-tracking Typhoon #22W
Nalgae made landfall in northern Luzon as a more compact but stronger
Category 4 storm (67 m s-1 sustained) on 1 October. Detailed discussion
of these storms can be found in the Joint Typhoon Warning Center Annual
Tropical Cyclone Report
(http://www.usno.navy.mil/NOOC/nmfc-ph/RSS/jtwc/atcr/2011atcr.pdf).
These three tropical storms changed the nature of the regional meteorology
for the second half of the cruise, and as we discuss, modulated regional
aerosol loadings. Satellite imagery clearly showed the region oscillating
between significant convection, developing in inflow arms (e.g., 22 and 27 September) across the SCS/ES, followed by areas of considerable clearing
(e.g., 24–25 and 29–30 September). Inflow arms corresponded with increases in
southwesterly winds, perhaps further ventilating MC air into the SCS/ES
region.
Evolution of the overall aerosol environment during the Vasco cruise period
To provide context to regional fire and aerosol behavior during the Vasco cruise,
time series of fire activity and AOTs are given in Fig. 4. Figure 4a
shows the MODIS fire hotspot time series for key regions in the MC for the
2012 burning season. As explained in Reid et al. (2012) to account for satellite orbit,
some smoothing of the data are required; in this case a 5-day boxcar is
used. Four fire events are visible over the course of the SWM. First, an
early-season event in late July/early August is visible in Central Sumatra
and Indonesian Kalimantan (predominately western Kalimantan); this is
associated with early agricultural burning. A second and much more
significant peak in late August is found in Southern Sumatra and Indonesian
Kalimantan provinces predominately in the south. This is fairly anomalous
behavior, especially for a La Niña year, as this region typically burns very
late in the season (Reid et al., 2012).
Contextual aerosol data for the 2011 aerosol season. (a) Combined
MODIS active fire hotspot prevalence by region. Data is smoothed in a 5-day
boxcar filter to help account for orbit. (b–e). Level 2 AERONET 500 nm
fine mode AOTs for key sites in the Southeast Asian region (marked on Fig. 2b). (f–h) Combined MODIS 7 MISR satellite AOT analysis for the early,
mid and late phases of the cruise.
In September, two more events, one early and one late in the month, are
visible. The first, peaking around 7 September is region wide, but is
dominated by Sumatra. The last major event, which corresponded with the
Vasco cruise, peaked 26 September, with major contributions from southern
Sumatra and Kalimantan and more minor contributions from islands to the
south of Borneo. As noted in Reid et al. (2012), these peaks in observed fire activity
often correspond to dry MJO phases (e.g., 23 August, 26 September) or overall weak
MJO activity (e.g., 5 September). The period of 20 July–8 August corresponded
with a late-phase MJO event. A new MJO event formed on 18 August. We suspect
drying ahead of the convective portion of the event perhaps allowed southern
Kalimantan to burn more readily on 23 August. The wettest phase of
the MJO (phase 3) was in the MC from 28 August to 18 September. A break in
precipitation in the southern MC allowed the 8 September fire event, which
was dominated by southern Sumatra, and the border of more significant
precipitation to the north. It is emphasized, however, that while we believe
plots such as Fig. 4a are indicative of qualitative fire patterns, they
are nevertheless influenced by clear sky bias, which also corresponds with
MJO activity.
While the MC generally has high background aerosol concentrations from
pervasive industrial, shipping and biofuel sources (Reid et al., 2013), peaks in AOTs
from AERONET sites largely match fire activity. Fine-mode AOT from four
sites are shown in Fig. 4: (b) Singapore; (c) Songkhla (further up the
Malay Peninsula in peninsular Thailand), (d) Kuching in Sarawak Malaysia,
Borneo, and (e) Notre Dame of Marbel University on Mindanao. Fine mode AOTs
from sites near sources typically ranged from 0.1 to 0.3 during background
conditions, and 0.4–1.0 during biomass burning events. For the most part,
the 23 August event was the largest region wide, with significant
spikes in both Singapore (impacted from Sumatra) and Kuching (impacted
largely by southern Kalimantan). The 7 September event is also
visible in Singapore, but there is little indication of smoke over Kuching.
The Vasco cruise period captured the last AERONET AOT peaks for the season in
Singapore, Kuching and in particular Mindanao. This establishes that the
ship was well positioned as a long-range receptor for transport from the MC
into the SW monsoonal trough.
Because of the generally small fraction of clear sky, frequent high thin
clouds, and sometimes extreme AOTs in the region, it is difficult to apply
satellite AOT retrievals in a straightforward manner. In particular,
sampling bias can be pervasive (Zhang and Reid, 2009). However, the AOT analyses in Fig. 4 that are associated with the meteorological modes presented in Sect. 3.2
are illustrative of regional aerosol loadings: (f) MJO active phase: 17–22 September; (g) MJO transition and TC active phase: 23–27 September; and (h) post TC
environment and clearing: 28–30 September. These AOT maps, coupled with the
large-scale flow patterns shown in Figs. 2 and 3, are suggestive of a
large-scale southwesterly transport event from the MC to the SCS/ES region
in the latter half of the cruise. Early in the cruise, while burning was at
a minimum, moderate AOTs still existed in the vicinity of Sumatra and
Borneo. Air was relatively clean north of the equator. During the
development of the TC active phase, the accelerated burning resulted in a 2–3 factor increase in observed AOTs in the source regions. Smoke
being transported into the SCS/ES, Celebes Sea, and Sulu Sea is clearly
visible. Due to clearing in the post TC phase, retrievals were then possible
over much of the region. Heavy smoke is observed as far as 10∘ N, with
moderate AOTs extending past Luzon. Cleaner air masses with AOT < 0.125 are clearly visible on the western side of the Philippines. Thus, from
the time series in both Figs. 2 and 4, we would expect aerosol
concentrations to increase as air masses entered the convective regions of
the SCS/ES. As no satellite retrievals were ever made on the track of the
Vasco, a question remains as to the aerosol concentrations within the active
regions. This is addressed in the next section where we discuss
environmental time series from the Vasco.
From an aerosol modeling perspective, Fig. 5 presents a time series of
AOT, surface anthropogenic fine-mode concentrations, and biomass burning
provided by the NAAPS reanalysis for key transitional days. Through use of
AOT data assimilation and satellite precipitation to constrain wet
deposition, this is a reliable global model scale perspective of aerosol
transport in this data sparse region. Shown are 4 of the days in Fig. 3: 18, 22, 24, and 30 September. By and large, modeled aerosol fields match our
expectations from the meteorology. While AOTs are high near source areas in
the first half of the cruise, convection over the SCS/ES quickly scavenged
aerosol particles near shore. This was particularly true for periods with
well-established TC inflow arms. In the second half of the cruise, two
strong injection and transport events carried aerosol particles as far north
as Luzon. These events were separated by TC Nesat. The relative strengths of
anthropogenic pollution vs. biomass burning suggest significant burning
enhancement in the last days of the cruise. Of particular note is that in
the middle portion of the cruise, model and flow data suggest the northern
Palawan region was most dominated by transport up the SCS/ES from the Java
Sea and Southeastern Borneo, with the Sulu Sea being dominated by transport
from eastern Borneo through the Celebes Sea. This Sulu Sea flow pattern then
dominated for the last few days of the cruise, although as discussed in the
next section, we suspect some additional industrial sources in the final
day.
NAAPS 550 nm aerosol optical thickness (AOT) and surface
concentrations for fine mode anthropogenic and biomass burning particle
concentrations for 4 key days during the cruise. Satellite data for these
4 days is also presented in Fig. 3. Cross sectional lines for Fig. 6
(24 and 30 September) are placed on the AOT plot.
Finally, aerosol vertical distribution is a crucial element of the system.
Unfortunately, CALIPSO was placed in standby mode from 22 to 30 September due to
solar flare activity. For the early cruise (17–22 September) thick regional
cirrus cover and orbital track conspired to prevent meaningful aerosol data
collections. However, the NAAPS reanalysis does provide a simulation of
aerosol vertical distribution, and we checked for consistency once CALIPSO
data was made available for 1 October when cirrus optical thickness was
low enough to profile the aerosol layers underneath. These data are
presented in Fig. 6. Meridional cross sections for total fine-mode aerosol
particle concentration are provided for 24 and 30 September, for 110 and
120∘ E longitude across the SCS/ES and Sulu Sea regions. These meridians
are marked on the AOT plots of Fig. 5. At Borneo and immediate outflow
regions, NAAPS generally keeps the bulk of the aerosol mass concentration
below 3 km, in line with previous remote sensing (Tosca et al., 2011; Campbell et al., 2013) and higher
resolution modeling efforts and comparison (Wang et al., 2013). We can interpret this as
smoke mixing though a deep planetary boundary layer (PBL), including its cloud
entrainment zone. This deep layer progresses well offshore east of Borneo in
the Celebes Sea. However, as we go further into the SCS/ES and Sulu Sea,
fine-mode aerosol particles concentrations are increasingly predominant in
the lowest kilometer.
(a–d) Meridional cross sections at 110 and 120 east of NAAPS
reanalysis total fine mode aerosol particle concentration for the 25 September (a, c)
and 29 September (b, d) haze events. (e) CALIOP 532
nm backscatter across the SCS/ES region on 1 October 2011. (f) Rescaling of
(e) for the lowest 4 km. Included is a map of the CALIPSO track.
Cloud-Aerosol LIdar with Orthogonal Polarization (CALIOP) data in Fig. 6, collected on 1 October 2011 (the day after the
Vasco returned to port but still probably representative of the second large
event), shows the same features, with perhaps an aerosol layer aloft at 1–2 km in northwestern Borneo, but a sharp aerosol layer below 1 km across the
SCS/ES region. In this case, the scale heights are even lower than NAAPS,
perhaps due to numerical diffusion in the vertical in the model. This
regional transition from deeper to shallower aerosol scale height, as one
moves out in the SCS/ES, is seen very clearly in climatological lidar data
(e.g., Campbell et al., 2013). In the context of this cruise, we can explain it as a result
of the veering wind shear in the lowest portion of the atmosphere. Aerosol
particles in the marine boundary layer (MBL) are transported with a more southwesterly wind. At 850 hPa and above, winds are more westerly. Thus, aerosol particles at higher
levels are transported eastward rather than north. Similarly, convective
lofting into the lower troposphere will then place the aerosol particles in
a westerly wind, and thus any northward component of transport must be
associated with the MBL. This finding makes understanding the sea breeze
induced ejection of smoke on the western side of Borneo all the more
important in the simulation of smoke transport to the Philippines and the
monsoonal trough. For eastward transport off of eastern Borneo, the boundary
layer and lower free troposphere winds have similar directions. Hence, we
find deeper aerosol layers in the Celebes Sea. Based on the climatological
aspects of wind shear (e.g., Reid et al., 2012), we expect this generally explains the
climatological aerosol vertical distribution in the region presented by
Campbell et al. (2013). This finding also suggests that the surface sampling by the
Vasco was largely indicative of smoke and pollution transport, and is
representative.
Results II: Vasco meteorology and aerosol time series
As Sect. 3 has established the overall nature of the lower troposphere, we
can begin to interpret the measurement time series from the Vasco. In particular,
we wish to understand how the large-scale conceptual models and observations
presented above relate to real world marine boundary layer meteorology and
aerosol phenomena. Key meteorological and aerosol measurements, which best
depict the overall environment, are presented in Fig. 7. Included are the
meteorological parameters: (a) pressure; (b) temperature; (c) wind speed;
and (d) precipitation rate. Key aerosol parameters include (e) the 30 min
average water CPC total aerosol concentration; (f) the estimated PM2.5 mass concentrations from filters (corrected to remove sea salt by
subtracting sea salt based on 3.26* Na concentration) and organic and black
carbon from quartz filters. Also shown are grab-can samples of CO; (g)
PM1 ammonium sulfate (NH4)2SO4 in red (based on DRUM
sampler S assuming all non-sea salt S was in (NH4)2SO4) with
coarse-mode sea salt in blue (1–10 µm, based on the 3.26* Na method),
and finally (h) NAAPS-derived total fine-mode particle concentration,
differentiated between biomass burning and a combined interactive
anthropogenic + biogenic product.
Marked on Fig. 7 are points of interest during the cruise to be discussed
herein. They begin with departure from Manila Harbor, followed by our exit
from Manila Bay. Our first point of stationary sampling was at Apo Reef,
followed the next day at the West Coron site. Long time-period stationary
sampling was then conducted at Guntao Island just outside of El Nido, then
just outside Malampaya Sound, and then back at Guntao Island again. During
the last Guntao Island measurement period, the Vasco experienced the largest cold
pool event, a topic of discussion of Sect. 4.2. Late on 26 September, the
Vasco took shelter from Typhoon Nesat in Liminangcong harbor, which showed
considerable local contamination. Once there was suitable reduction in
significant wave heights, the Vasco moved north to just outside El Nido harbor to
enable more regional sampling. On the morning of 29 September, the Vasco had to
return to Manila harbor via the Mindoro Strait ahead of TC Nalgae. In
preparation for Nalgae, our equipment was shut down and boxed up one-third
of the way into Manila Bay midday on 30 September.
Cruise time series of key meteorological, aerosol and chemistry
indicators in 1 min intervals. Key sampling points and events are marked
by vertical lines. (a) Surface pressure (hPa); (b) ambient air temperature
(∘C); (c) wind speed (m s-1); (d) precipitation rate (cm h-1);
(e) CPC total particle count; (f) left axis: PM2.5 gravimetric
mass with sea salt subtracted, and associated organic and black carbon;
right axis – dots: can carbon monoxide (ppbv); (g) left axis – red: DRUM
impactor time series of inferred PM1 inferred ammonium sulfate (µg m-3); right axis – blue: inferred coarse mode sea salt
(dp > 0.8 µm). (h) NAAPS total fine mode particle mass
segregated into anthropogenic (+biogenic) fine mode and biomass burning.
Based on a preliminary analysis of NAAPS data (e.g., Fig. 5), boundary
layer air sources were all coastal Borneo or Southern Sumatra/Java Sea for
most of the cruise. The two important exceptions were in the first day and
last 2 days of the cruise, when model and trajectories suggest some
influence from northern Borneo and the Celebes Sea. As discussed above, winds
veered with height, with the lower free-tropospheric air tracing an origin
to the Malay Peninsula and Indian Ocean, where pollution and biomass burning
emissions are significantly reduced. Thus, we expect highest particle
concentrations to be in the MBL.
Daily scale meteorological and aerosol concentration features
To understand the nature of the coupled meteorological-aerosol environment
we have to reconcile large scale meteorological and remote sensing analyses
with the data at a single receptor point (i.e., Vasco). Clearly from Fig. 7,
both the meteorology and atmospheric composition observed on the cruise are
a convolution of low to high frequency signals. To begin the analysis, we
consider features with scale of a day or longer.
As we would expect for a tropical region, overall we see a large measure of
consistency in many meteorological features. At daily scales, pressure is
relatively constant for the cruise with the exception of a moderate dip
∼ 26–29 September associated with TC Nesat and an embedded
diel-solar-tidal signal. Baseline temperatures are also constant at
∼ 28 ∘C, with a 2 ∘C dip also associated with heavy
rains from the TC. Surface winds were generally 5–10 m s-1 and
typically from the Southwest with occasional departure to the north.
Precipitation was showery throughout, with precipitation visible in some
form most days, but with the most significant events in the outer rain bands
associated with TC Nesat. Embedded in these daily scale features are clear
high-frequency phenomena; for example, inverse ramp drops in temperature,
with associated spikes in wind speed, and often precipitation. As discussed
in Sect. 4.2, such high-frequency phenomena are largely associated with
convective cells and their associated cold pools.
Within the cruise, we see several large-scale aerosol features. Certainly,
just before the Vasco left Manila Harbor and Bay, we observed a high spike in
particulate matter, indicative of local pollution. However as the Vasco departed,
we entered a cleaner greater-bay regime, upwind of Manila Bay sources.
Outside of Manila Bay, a spike in particulate matter was also observed,
likely due to local Luzon influence such as from Batangas. “Regional”
SCS/ES monitoring was initiated with the Vasco's first anchorage at Apo Reef in
Mindoro Strait on September 18. A more typical background period was
observed through midday 22 September, followed by a significant aerosol
event ∼ 24–26 September ended by the arrival of TC
Nesat. A second even larger event then followed from late 28 September through
the return on 30 September.
From the Apo Reef to the northern Palawan anchorages on 23 September ,
the Vasco was in a very clean aerosol regime. CN counts were generally on the
order of ∼ 300–500 cm-3, and non-sea salt PM2.5 was ∼ < 2 µg m-3. PM10 sea salt was
on the order of 5 µg m-3. Both fine and coarse particle mass are
in line with expectations in a background marine atmosphere (Quinn et al.,
1996; Henintzenberg et al., 2000; Reid et al., 2006). On 22 September,
particle concentrations reached a mission minimum, with sustained CN
concentrations below 150 cm-3, and non-sea salt PM2.5 < 1 µg m-3; at or below our minimum detectable limits. Coarse-mode sea
salt remained relatively constant, increasing slightly to 6 µg m-3. During this time period, however, we found variable CO grab sample
data ranging from 80–118 ppbv, uncorrelated with particle properties. This
first period can be explained through the development of TC Haitang near the
SCS/ES, and the formation of a broad southwest to northeast inflow arm on
22 September clearly visible in Fig. 3. As the inflow arm developed, winds
accelerated and precipitation from both shallow and deep convective cells
increased. Thus, while Borneo/Java Sea air was clearly being transported to
the Vasco receptor, precipitation scavenged most fine particles, leaving
insoluble trace gases but few particles. Pulses of slightly enhanced CO
nevertheless reached the ship. NAAPS correctly captures this period as
relatively clean, although total mass concentrations are high by
∼ 2–3 µg m-3.
The first observed regional aerosol event having a clear Indonesian or Malay
source was initiated on 23 September, when Haitang moved westward, leaving
clearer skies and lighter winds. The Vasco remained at the same anchorage outside
of El Nido for this entire event. This period saw a slow development in
particle concentrations and CO and was largely precipitation free. Non-sea
salt PM2.5 was on average 8–9 µg m-3, with black carbon and
organic carbon mass fractions on the order of 5 and 20 %, respectively.
Corresponding CN counts were on the order of 1000–2000 cm-3. This
period also corresponded with reduced surface winds across the SCS/ES, and
an associated slight reduction in coarse-mode sea salt. A significant dip in
particle concentrations and temperature was observed late 24 September UTC
(∼ 3 a.m. local time), which, as we discuss in Sect. 4.2, was
associated with a strong trans-SCS/ES convection-cold pool event. Finally,
fine particle mass concentrations reached a maximum and then fell
precipitously with the arrival of storm conditions associated with TC Nesat.
NAAPS identified this event well as a mixture of anthropogenic and biomass
burning sources, although total fine-mode mass concentration is
overrepresented by ∼ 30 %. We suspect this is a result of a
low bias in the NOGAPS RH field, which in the context of AOT data
assimilation well upstream of the Vasco, results in an overestimation of dry mass
relative to ambient scattering.
During the storm period, the Vasco was in safe harbor at Liminangcong; the high
and variable CN are due to local harbor emissions. After TC Nesat passed,
the Vasco returned to El Nido for a day of measurements and eventual departure
back to Manila. This cruise return period was associated with very light
winds and the highest observed particle concentrations, perhaps with a
Borneo source. Again, such fair weather is expected on the back side of a
strong tropical cyclone such as Nesat, and was further reinforced with the
impending arrival of another Category 4 storm, TC Nalgae (Fig. 3, 30 September).
Fortunately, the typical southwesterly winds slackened to such an extent
that the ships own velocity kept air moving over the bow, thus avoiding
self-sampling that would have ruined the return period data set. A
time-series analysis of model and trajectory shows that, leading up to this
event, transport associated with the last vestiges of the TC Nesat's
influence in accelerating regional winds brought the air mass up to the
sampling region. Due to wind shear, it is possible it included contributions
from both western and eastern Borneo. While we cannot dismiss the
possibility of local contamination in the gas can samples while we were in
safe harbor in Liminangcong, we do see a steady increase in CO reaching a
plateau during the final event.
As the Vasco left El Nido, black and organic carbon mass fractions were on the
order of 5 and 40 % suggestive of biomass burning dominance. This
period also afforded the only cirrus-free conditions for Microtops sun
photometry measurements. 500 nm AOTs were on the order of 0.30, very similar
to the MODIS retrievals shown in Fig. 4h. NAAPS also captured this event
well, and yielded a correct 0.3 AOT. However, like the previous event, total
particle concentrations are biased high. Again, we suspect this is due to a
low bias in the NOGAPS RH fields. Even so, NAAPS suggests a significant
enhancement in biomass burning particle concentrations relative to
anthropogenic pollution.
Based on back trajectories, NAAPS simulations, and particle concentrations,
one would initially be inclined to believe the Vasco sampled one air mass on its
return to Manila. However, examination of wind data shows westerly to
northerly winds at the very end of the mission. This plus chemistry (Sect. 4.4 and Lagrosas et al., 2015) show that in the last 6 h of the
cruise there are slight perturbations to the sources, perhaps a change in
the mixture of biomass burning and industrial pollution or the addition of a
regional shipping signal. Indeed, across the horizon on 30 September we saw many
high polluting vessels with plumes visible from 10 to 30 km away.
A final consideration for large scale observations is how aerosol loading
covaries with atmospheric soundings, perhaps influencing interpretation of
aerosol, cloud and precipitation interaction studies. Figure 8 presents
three example cases were we found isolated convection, 18, 25 and 29 September. 18 September was our first stop at Apo Reef, where we
observed relatively clean aerosol conditions and isolated convection. Over
the 24 h period we observed many warm rain events with
significant precipitation, as shown in Figs. 8a and 7e. For
intermediate pollution on 25 September, we encountered significant amounts
of boundary layer clouds, but little precipitation (Figs. 8b and 7d). On the other end of the spectrum, 29 September was indicative of
polluted conditions where there were few boundary layer clouds, but
occasional significant convection (Figs. 8c and 7d). Simple
correlation studies and current scientific thinking would suggest these
cases epitomized aerosol–cloud–precipitation interactions. That is, in clean
conditions, we have significant amounts of warm rain. If aerosol particle
concentrations are perturbed from background conditions, warm rain ceases,
and perhaps there is enhancement in severe cells. However, as demonstrated
in Fig. 8d–f, atmospheric soundings were very different for these
cases. Being the tropics, one expects relatively consistent
potential temperature profiles, which indeed we found to be largely the case
(Fig. 8d). But, we can see that for the polluted 25 September case, a
clear stronger inversion is present at 700 hPa. This inversion corresponds
with a lower free-tropospheric dry layer between 900 and 700 hPa with both
halved water vapor mixing ratio (Fig. 8e) and relative humidity (Fig. 8f). This certainly impaired the development of warm rain formation, even
without possible aerosol effects. For the most significant biomass burning
event (29 September), the PBL was drier than was typical, yet the lower
troposphere was relatively moist. But in this case, large TC induced
subsidence produced a dry layer in the mid to upper troposphere, strongly
capping convection.
Photographs and corresponding sounding elements for three aerosol
regimes during periods of marginal convection. (a) 18 September at Apo
reef with isolated warm convection in moderately moist conditions; (b) 25 September at El Nido with warm non precipitating convection with a lower
troposphere dry intrusion during the height of the pollution event;
(c) 29 September at the northern Sulu Sea with isolated deep convection in
overall TC induced subsidence during height of biomass burning event. (d–f) Corresponding Vasco released radiosonde profiles of potential
temperature, water vapor mixing ratio, and relative humidity, respectively.
To better understand the nature of dry stable layers, Fig. 9a, b
present back trajectories initiated at the key “dry altitudes” of 1.6 km
(850 hPa) and 6.8 km (500 hPa), respectively, for our cases of 18, 25
and 29 September. Tick marks are located every 24 h, and time–height
dependencies are provided. For the lower free troposphere, we see clear
differences between 18 and 25 and 29 September, with 18 September originating from convection off of Borneo. For both the
25 and 29 September, the lower-to-middle free tropospheric air
originated in the Indian Ocean. The NOGAPS time–height cross section over
the Phuket, Thailand radiosonde site clearly shows a dry air intrusion into
the region between 2 and 5 km (900 and 600 hPa). This may be related to
subsidence behind the propagating MJO. Nevertheless, it does demonstrate how
dynamics in the Indian Ocean and the formation of dry layers can be coupled
to SCS/ES and Sulu Sea convection and their aerosol environment. In regard
to upper-level subsidence, trajectories are highly divergent, but show
significant lifting and subsidence associated with the passage of TCs.
Back trajectories and time height cross sections. (a, b) 1.6 km and 6.8 km back trajectories from the Vasco for the cases posted in Fig. 11.
(c) Time height cross section for Phuket, Thailand, of relative
humidity color (RH) with potential temperature isopleths (∘C). Wind
barbs are given with full and half bar at 10 and 5 m s-1, respectively.
High frequency squall line and cold pools phenomenon
Embedded in the Fig. 7 time series are clear, sharp perturbations in both
meteorological and aerosol features. Most significant of these are drops in
temperature on the order of 2–5 ∘C within minutes, and even here we
must consider the response time of the aspirated temperature probe. With the
drop in temperature, there was a sharp spike in wind speed, relative
humidity (and at times precipitation), as well as a drop in both particle concentration
and water vapor mixing ratio. These characteristics are indicative cold pool
events related to convective downdrafts (Wakimoto, 1985; Atkins and Wakimoto, 1991; Miller et al., 2008; Zuidema et al., 2012). Over
20 such events are observable in the time series, with significant
variability in amplitude. Recovery from the drop in temperature and particle
concentration to the pre-event baseline ranged from 1 to 10 h. Some
of these events originated from what were clearly local isolated cells.
However, investigation of the largest such events suggest that they
originate in long-lived squall lines, propagating in the monsoonal flow and
initiated from the cold pools of massive thunderstorms over land or along
the coast. This phenomenon appears to be extremely important for determining
aerosol fate in this region, and deserves detailed study in its own right.
For this study, we will limit our discussion to the most significant event
observed during the cruise.
24 h times series of meteorology and aerosol
parameters centered on the 24 September cold pool event. Tines are in
UTC. (a) 1 min temperature and wind speed; (b) 1 min relative humidity
and pressure; (c) PCASP and CPC total aerosol particle count; (d) and
(e) PCASP number and volume distributions, respectively.
The pathology of SCS/ES organized squall line/cold pool phenomena best
described by the cruise data was for a 24 September event in the middle of
the first significant aerosol transport episode. Key aspects of the 24 September event are presented in Fig. 10 as 1-minute averages. Included
are (a) a time series of temperature and wind speed; (b) relative humidity
and pressure; (c) PCASP and CPC total particle count; and PCASP (d) number
and (e) volume size distributions. The cold pool hit at 16:28 UTC
(corresponding to 00:28 LST on 25 September). Wind cup speed accelerated from
the background 7–8 m s-1 to 14 m s-1 within the first 2 s,
with flux estimates of gusts at the 2–5 second level to 25 m s-1 within
the next 50 s. Winds then momentarily subsided to 5 m s-1 for the next 10 min, followed by another increase and
decrease over the next hour, and a slow recovery. Corresponding with the
wind onset was a ∼ 5 ∘C drop in temperature, and increase
in relative humidity over the first minutes, although there was only a minor
0.2 hPa perturbation in pressure. Sea surface temperature dropped 0.2 ∘C
and recovered only after sunrise. Approximately 1 cm of precipitation
occurred over a 1 h period, which initiated 15 min after gust front
arrival, thus breaking the wind lull. Maximum precipitation rate was on the order
of 4 cm h-1. Surface particle concentrations dropped precipitously
with cold pool arrival: PCASP counts dropping from ∼ 700 cm-3 to 300 cm-1 within 2 min, followed by a further
reduction to 150 cm-1 at precipitation onset. CPC dropped from
∼ 1450 to 400 cm-1. An interesting feature was a clear
enhancement in coarse-mode sea salt along the gust front. This is, to our
knowledge, a first ever report of a maritime corollary to dust producing
haboobs (Knippertz et al., 2007; Miller et al., 2008; Seigel and van den
Heever, 2012). Particles and meteorological parameters likewise recovered to
pre-event levels over the next 10 h.
While the 24 September event was the largest of its kind, it nevertheless
demonstrated patterns similar to over 20 other events: a sharp wind
increase and temperature and particle decrease is followed by a lull and
eventually precipitation from a cell. When these events occurred in
association with isolated cells, we often could observe the entire process
from cell formation to cold pool onset and, at times, cell propagation over
the site. Investigation of the 24 September case, however, led us to a
conclusion that despite the short spatial and temporal timescales observed
at a receptor site such as the Vasco, they are part of a meteorological
phenomenon that spans the entire SCS/ES region. Visible and IR satellite
imagery of the SCS/ES region for the 18 h prior to the 24 September event are presented in Fig. 11. At arrival, the cell was only
30–50 km along the meridian, with cloud top heights on the order of 12–13 km, well below the 18 km tropopause height. Tracing the event back in time
with 15 min imagery, we found this system, despite its small size,
remained organized for nearly 24 h. Imagery suggests that an
isolated thunderstorm that formed near the southern tip of Vietnam/Ho Chi
Min City initiated a cold pool southward which eventually embedded within
the Southwest monsoonal flow. This cold pool triggered an arc cloud
formation that triggered a new set of thunderstorms along the arc, which in
turn formed a secondary cold pool and repeated.
Day visible and night infrared time series of 24 September
squall line/cold pool event. (a) 24 September 0Z NOGAPS surface and 700 hPa winds at event initiation. (b) 23 September 14:32 UTC cold pool arc
cloud propagating south from Ho Chi Min City initiated thunderstorm.
(c) 24 September 00:32 UTC, convective cell spawned by cold pool, propagating to
the NNE; (d) 24 September 06:32 UTC cold pool from cell in (c);
(e) convective cell spawned by cell in (e); (f) final cell spawned by cold pool
from (e) sampled by Vasco; (g, h) 250 m MODIS Aqua Ch 1 visible and
derived cloud height product respectively. Inset in (d) is the domain.
(i) 18 September 01:32 UTC MTSAT image of extensive latitudinal dimension of two squall
line events.
Squall line features such as observed here have been long noted in the
literature (e.g., Trier et al., 1996), although we have been unable to find cases as
long-lived as we found during the cruise. There are some similarities in the
radar science literature for mid-latitude systems as “bow echoes”
(Weisman, 1993). The physics have been studied extensively (e.g., Weisman and Rotunno, 2004), and the
importance of vertical wind shear and the presence of mid-tropospheric dry
air behind the storm front is well established. However, the nature of the
squall lines in the SCS/ES appears to present an extreme case. Figure 11g and h show the MODIS Aqua 670 nm visible and cloud top height products for
the 24 September event, 10 h before it reached the Vasco. Shown is a pair
of squall lines, with the southern arc being the one that eventually
developed most strongly. We find it interesting that, for the most part, the
tops of the clouds making up the squall lines reached only 5–6 km, and hence
were most likely ice-free. Only isolated cells along the arc became high
enough for freezing and further vertical development. However, a review of
the satellite loop suggests periodic major storm eruptions along the line,
which we surmise help propagate the phenomenon. In comparison, classic
mid-latitude bow echoes are very deep along the front; the difference in
cloud heights may be related to the relatively larger amounts of CAPE aloft
in mid-latitude systems (Takemi, 2014), as well as the location of the capping
inversion. Long-lived squall lines are known to develop in environments with
finely tuned balance between shear and CAPE (Rotunno et al., 1988). The
question of whether cold pool propagation is drive by the frequent and
relatively shallow convection or the infrequent troposphere-deep convection
is one we plan to study in detail in the near future. From an aerosol point
of view, the warm versus cold convective components along the line likely
have important ramifications for scavenging or redistribution of aerosol
particles in the MBL. Similarly, aerosol impacts on warm versus cold
convection are likely different. Aerosol particles have even been
hypothesized to influence the cold pools themselves (Lebo and Morrison, 2014).
A second important aspect of these cold pools is their extent across the
monsoonal flow. The case experienced by the Vasco, while long-lasting, was
relatively small in dimension. Frequently, much larger events are observed
in our analysis of the satellite data record. An example at the beginning of
the research cruise (18 September) is presented in Fig. 11i. In this case,
younger and more developed squall lines are shown, each over 500 km in
length. These events were initiated by major thunderstorms over and just
offshore of the Malay Peninsula, with overshooting tops of > 20 km. They propagated across the entirety of the SCS/ES in under 30 h.
With such wide ranging extent, they must have swept across the entirety of
the SCS/ES, perhaps leaving the very clean condition observed in the
northern area. Imagery analysis showed the southern portions of these squall
lines developing more strongly on their southern half. This suggests that
indeed the veering wind shear is supplying energy from the southern domain.
Key aspects of chemistry and particle microphysics
Detailed analysis of aerosol chemistry, size, and optical properties will be
presented in subsequent papers. However, there are key aspects of chemistry
and size worth briefly discussing in the context of this regional aerosol
source and transport paper. Time series of DRUM sampler derived PM1 for
some key elements are presented in Fig. 12: (a) sulfur and potassium and
(b) aluminum and vanadium, respectively. Key gas species of CO and benzene
are presented in Fig. 12c as is 2-PenONO2 (a photo-oxidation
product for pentane) and methyl iodide (CH3I) a marker for biomass
burning (Akagi et al., 2011). While aerosol source identification in the complex
Southeast Asian environment can be very involved (see e.g., Atwood et al.,
2013a, b), there are significant features of note. First, though non-sea-salt
sulfur can be produced by both industrial and biomass burning (particularly
peat burning for sulfur), potassium shows significant enrichment during
flaming biomass burning (Reid et al., 2005; Akagi et al., 2011).
Time series of key elements and gases. (a, b) DRUM time
series of sulfur + potassium and aluminum + vanadium, respectively.
(c) Carbon monoxide and benzene, both common biomass burning emissions.
(d) 2-Pentane oxyl nitrate, a photochemical pentane daughter product and
methyl-iodide, a halogenated organic specie also emitted by burning, the
oceans, and used in agriculture.
Particle size and chemistry characteristics for aerosol environments
at the cruises's key sampling locations.
Date
Sample location
Suspected
Mode:CMD:σgn
Mode:VMD:σgv
BC % / OC %
K/S
source
(µm, µm, n / a)
(µm, µm, N / A)
16 Sep
Manila Harbor
Metro Manila
12 % / 19 %
0.01
17 Sep
Manila Bay
Local Bay
0.17 : 0.16 : 1.73
0.285 : 0.30 : 1.43
0.02
17 Sep
Outside Manila Bay
Sulu Sea/N. Borneo
0.11/0.17 : 0.13 : 1.37
0.19 : 0.21 : 1.52
Bdl / 28 %
0.08
23 Sep
Malampaya Sound
Malay Pen. & Sumatra
N / A
N / A
2 % / 58 %
0.12
25 Sep
El Nido
SW Borneo
0.17 : 0.17 : 1.61
0.285 : 0.27 : 1.36
5 % / 27 %
0.10
29 Sep
N. El Nido
Southern Borneo
0.24 : 0.20 : 1.54
0.31 : 0.29 : 1.28
5 % / 30 %
0.29
30 Sep
Outside Manila Bay
N. Malay Pen. through
0.17 : 0.18 : 1.56
0.31:0.28:1.31
7 % / 31 %
0.23
Vietnam
By and large, sulfur and potassium track with each other over the time
period, with a significant enrichment in the post-TC Nesat clear area.
Aluminum, indicative of regional fine dust, or at times fly ash, also tracks
sulfur well and potassium quite well, perhaps indicative of soils entrained
in biomass burning plumes. As an indicator of industrial or oil combustion,
vanadium shows two significant spikes on 26 and 30 September. This may
indicate additional industrial or shipping sources. Based on our trajectory
analysis, these cases may very well be influenced from the industrial
Singapore–Kuala Lumpur corridor, although high-resolution modeling is
required to show this with any certainty. From a gas chemistry point of
view, we find that fine aerosol and CO match reasonably well, with the CO
enrichment ahead of the 24–26 September aerosol event perhaps indicative of
polluted air masses where particles have been scavenged by precipitation.
Benzene, a good and relatively stable indicator of biomass burning and some
industrial emissions, also tracks CO, though with perhaps less enrichment in
the last day of the cruise. Methyl iodide tracks with potassium as we would
expect from a biomass burning tracer. As 2-PenONO2 is a photo oxidation
product, its presence demonstrates that these plumes are nominally well
aged, particularly for the first event. A reduction in 2-PenONO2 for
the last day of the cruise with an enhancement vanadium suggests a change in
air mass sources and/or aging. At the same time, the ratio of ethyne to
excess CO can also be used as a photochemical clock for plume aging. While
relatively noisy from the cruise, it ranged from 15 for the 18 September
spike, suggesting a fresh source, was consistently lower (2 to 5) for the 28–30 September event suggested uniformity in fair degree of photochemical aging.
Conversely, the 24–26 September event showed more variability (3 to 8),
suggesting more mixed photochemical aging and perhaps sources. Such
chemistry must be further analyzed with the aid of numerical models.
Regarding aerosol size properties, fine-mode size distributions exhibited
some variability throughout the cruise (Fig. 12; Table 1). Number
distributions showed relatively strong trends, with cleaner periods having
significantly smaller count modal diameters (∼ 0.11 to 0.24),
though curve fits generally converged to count median diameters in the
0.13–0.17 range. Implicit in this is variability in geometric standard
deviation, which may have significance in regional aerosol–cloud
condensation nuclei studies. Also evident in the number distributions is a
frequent shoulder on the large side of the distribution, suggesting
differences in aerosol physics and chemistry for the number and volume
distributions; not uncommon in mixed environments. Volume median diameters
were generally in the 0.27–0.29 µm range for more polluted events,
further exhibiting larger overall size. Actual volume modal diameters are
slightly larger (∼ 0.02) than their curve-fit counterparts.
These are typical for both regional pollution and biomass burning
environments (Reid et al., 2005, 2013), and are comparable to the AERONET derived VMDs (Volume Median Diameter) by
Salinas et al. (2013) of 0.26–0.40 µm for background and severe smoke haze events and
the mean value of 0.32 µm by Reid et al. (2013) when one considers hygroscopicity.
An interesting aspect of the particle size and chemistry data for
high-frequency events is exemplified by the 24 September cold pool case.
Selected 30 min average volume distributions taken from the
1 min time series in Fig. 10e are presented in Fig. 13c.
Thirty min average volume distributions leading up to the cold pool
event, and 24 h later are nearly identical. In the 10 min
after arrival, we find a ∼ 80 % reduction in total particle
volume, with another factor of 2 reduction following the precipitation
event. All this time, VMDs remained fairly stable, although a clear increase
in larger particle concentrations is observed post wind burst. Between
Fig. 13c and Fig. 10d, e we do not see large changes in
particle size, but rather only in amplitude. Similarly, ratios of aerosol
chemistry are also fairly similar. We can interpret this data and the 7 h before initiation of aerosol population recovery as a sweep of clean
air aloft and subsequent further rainout of aerosol particles along the cold
front. Given the 3–4 m s-1 marine boundary layer wind speed, over 7 h we expect a roughly 75 km zone of marine boundary layer particles
being cleaned out by the event upstream of the Vasco. Such a length scale is
supported by the satellite images presented in Fig. 11, suggesting a
∼ 120–160 km swath was cut by the event.
PCASP size distributions for selected regimes. (a, b) Number and volume distributions for early, middle and late cruise periods.
(c) Volume distributions corresponding to the 24 September cold pool
event.
Discussion and implications for cloud and precipitation studies
This paper had three primary objectives: (1) provide a broad overview of the
2011 Vasco cruise, including instruments carried, cruise track, and the general
characteristics of the regional environment sampled; (2) relate how aerosol
properties co-varied with regional meteorological phenomenon and establish
the extent to which biomass burning or industrial pollution from the
southern maritime continent can be transported towards or into the boreal
summer southwest monsoonal trough; and (3) create a narrative based on field
data to help bridge climatological indicators commonly used to assess
aerosol life cycle to real world meteorology. To our knowledge, these are the
first published aerosol field measurements in the boreal summertime SCS/ES
region.
Central to all meteorological and atmospheric compositional questions for
the greater maritime continent is the role of convection. As discussed in
Reid et al. (2012, 2013), if ENSO-induced precipitation anomalies influence the overall
interannual variability of burning activity, it is the patterns of
convection correlated with MJO indices that best describe the specific
timing and lifetime of emissions. Indeed, the importance of the MJO to
meteorological phenomenon of the MC cannot be understated (Zhang, 2014). Yet we
understand very little of the mechanisms of MJO propagation across the
region. Embedded in the large scale “forest” point-of-view of ENSO,
monsoonal transitions, and the MJO are individual “trees” of specific
aerosol and convective events that can be quite diverse in nature, resulting
in complex relationships across land, ocean and atmospheric processes.
From the “forest” point of view, the Vasco observed aerosol and meteorology
phenomena that largely matched the conceptual model of MC aerosol
relationships between fire activity, transport and MJO transport put forth
in Reid et al. (2012). The entire 2011 burning season was represented by fire activity
slightly elevated with what one expects from a moderately cold ENSO year.
Timing of specific burning events was largely consistent with drier phases
of the MJO for the western MC (Phases 1 and 5–7). The cruise fortunately
took place during an MJO propagation from 3 into 6, and towards the end of a
significant burning event, and so sampled some very clean air as well as the
highest AOT recorded in the region for that season (Marbel University
Mindanao peaked at 500 nm AOT of 0.46, likely as a receptor for southern
Kalimantan burning on 28 September).
At the next level of scale, the migration of the MJO into phase 5 around
22 September coincided with the development of regional TCs, as described by
Maloney and Hartman (2001). This included the early-cruise development of a TC in the SCS/ES
and the pair of late cruise Category 4 TCs propagating westward across Luzon
at the very end of the mission. These TCs clearly enhanced convection along
a 2500 km inflow arm spanning the Sumatra/Malay Peninsula to Luzon, and yet
also are apparently associated with clear periods and rapid aerosol
transport. Indeed, the inflow arm that creates convection, and hence wet
deposition, can, at the end of its life cycle, perhaps rapidly carry more
polluted air masses into the SCS/ES and Sulu Seas. In these cases, smoke and
anthropogenic emissions from Sumatra and Borneo flowed deep into the greater
SCS/ES and Sulu Sea regions. It is quite possible that without TC influence,
such events would never have been observed. Control for TC activity is a likely necessity in any climatological analysis of regional aerosol transport.
At the finest scales, we were impressed by the nature of
coherently propagating squall line systems across the SCS/ES region, and how
these perhaps cut large swaths of aerosol particles out of the environment.
Even a cursory view of geostationary data in Fig. 11 shows how convection
moves along isolated lines embedded in the SCS/ES monsoonal flow. These
features are contrary to the more “bubbling pot” concept of tropical
convection in large-scale waves. Examining the entire mission data record,
we tracked dozens of lines of convection on the order of 100–500 km in
latitudinal length, propagating eastward. Cold pools of storms clearly
initiate new convection, which forms another set of cold pools and so on.
Veering wind shear allows these storms to cut across aerosol particles
transported in the marine boundary layer, effectively removing them from
that altitude regime. Perhaps the dry air intrusions in the lower free
troposphere from the Indian Ocean provides needed dry air to perpetuate the
bow echo-like form observed. But this is speculative at this time and much
more research is needed on the physics and conditions that support long
squall line phenomenon.
From an aerosol point of view, the prevalence of high-resolution features
like cold pools, and the warm versus cold convective components along the
line, likely have important ramifications for scavenging and/or
redistribution of aerosol particles in the MBL. Aerosol particles have even
been hypothesized to influence the cold pools themselves (Lebo and Morrison, 2014), offering
up a potential feedback. While there have been many attempts to correlate
convective activity with aerosol indicators, such as AOT, organized squall
line behavior such as presented here will defeat such a methodology. In the
24 September case, the high winds of the cold pool were ahead of the
precipitating cell. Thus, particle concentrations were dramatically reduced
before the cell arrived. In a study of the influence of cold pool generated
dust on the parent convective cell, Seigel and van den Heever (2012) found the dust had little
effect. Vertical transport of the dust was harmlessly ingested at
mid-levels. No doubt, the burst of sea salt produced by the cold pools
observed on the cruise would meet a similar fate. But, the findings of
Seigel and van den Heever (2012) have perhaps a more interesting corollary. If wind generated aerosol
particles do not have a significant effect, do the aerosol particles ahead
of the cold pool also have a lesser effect? Are these particles vertically
redistributed and eventually entrained into the clouds at mid-levels as
well? Finally, what then is the role of vertical wind shear in bringing
aerosol particles from the south into the squall line convection? These
questions on aerosol life cycle and impacts relate back to the convection
physics and the nature of clouds within the squall line. From Fig. 11h,
cloud tops along the squall line are at 6 km or above, but the efficiency of
aerosol scavenging by these features is unknown, although we suspect they
are important sinks for regional particles.
The strong relationships between convection patterns, emissions, and
transport have serious implications for regional study of aerosol impacts on
clouds and precipitation. Even more so, these process implications propagate
further into climate change projections. While the studies of Reid et al. (2012) and Xian et al. (2013) provide a good climatological foundation for
aerosol life cycle, they are nevertheless a substantial smoothing of highly
intricate ejection and convection interactions. However, just because
relationships are complex does not imply they are fundamentally chaotic.
While future papers will describe in more detail the covariance between
aerosol particles and convection, it is appropriate to close this paper
recalling the covariance between aerosol populations in the MBL and key
features in atmospheric soundings in Fig. 8. Indeed, the presence of
substantial amounts of smoke in the boundary layer is fully intertwined with
reduced convection and the presence of dry layers aloft, either through large
scale subsidence or dry air. At the same time, these dry layers likely
influence the gross type and structure of convection irrespective of aerosol
particles as CCN. In future studies, we will attempt to constrain aerosol
causality components from thermodynamic forcing of regional convection. At
the heart of such an endeavor is understanding what controls convective
initiation. Clearly, any aerosol–precipitation study has to account for such
complex meteorology. Then, when one considers the implications of
aerosol–precipitation feedbacks of a changing climate, we must consider how
such phenomenon as ENSO, monsoonal transitions, the MJO and TCs will
themselves change. For these phenomenon the community is already challenged
to perform medium range to seasonal forecasts, let alone develop consistent
simulations in climate models. Thus, perhaps the most important lesson of
this work is that all aerosol–climate interaction research for the region is
predicated on further advancements of fundamental meteorological processes.
Conclusions and hypotheses for future work
This paper provides a broad overview of the 2-week research cruise of the
Vasco for 17–30 September 2011 in the northern Palawan archipelago of the
Philippines. The ship was stationed on the windward side of the boreal
summertime southwest monsoonal trough, influenced by marine boundary layer
(MBL) air originating from the islands surrounding the Java Sea. Lower free
tropospheric air above the MBL largely originated in the Indian Ocean,
passing through and over the Malay Peninsula. Based on the analysis of Reid et al. (2012), we suspected this region's MBL is impacted by anthropogenic
pollution and biomass burning emissions from Indonesia, Malaysia, and
Singapore. Given Southeast Asia's ubiquitous cloud cover, it is difficult to
determine by remote sensing what the impact is of anthropogenic activities
on aerosol populations in a region suspected to be vulnerable to aerosol
impacts (Reid et al., 2013). What we do know is largely based on modeling studies, which
have difficulty with this most complex of meteorological environments.
Hence, this cruise provides the first ever, to our knowledge, contiguous
measurements of the South China Sea/East Sea (SCS/ES) and Sulu Sea aerosol
environment. Based on this cruise, and a subsequent 1-month September 2012
Vasco cruise to be reported on later, we observed enough of the environment to
study aerosol life cycle and pose questions for targeted analysis and testing
of cloud impacts. At the very least, the 2011 cruise provides a narrative of
real world meteorological phenomena to provide realistic conceptual models
of how the regional aerosol life cycle relates to the southwest monsoonal
system. In summary, we reported on the following:
Boreal summertime 2011 was an El Niño–Southern Oscillation (ENSO) cold
“La Niña” phase year, yet had slightly above-average burning activity for this
inter-seasonal state. While peak burning and aerosol optical thicknesses (AOTs)
on Sumatra and Borneo for 2011 occurred in mid-August, with > 0.8
fine mode 500 nm AOTs recorded by AERONET, the end of the Vasco cruise
corresponded to the largest aerosol injection into the Philippines, bringing 500 nm fine mode AOTs on the order of 0.3 to 0.4.
The Vasco cruise corresponded with Madden–Julian Oscillation (MJO)
propagation from phase 2 to 6, which should enhance burning and transport (Reid et al., 2012).
With MJO propagation came significant tropical cyclone (TC) activity, including
the formation of a tropical storm in the SCS/ES in the early part of the cruise
(Haitang), and the propagation of two Category 4 storms at the very end (Nesat and Nalgae).
This TC activity strongly modulated winds and convection in the greater SCS/ES and Sulu Sea, and thus aerosol regional transport and life cycle.
Active convective phases associated with TC development and inflow arms
demonstrated extraordinary clean conditions, with condensation particle counter
(CPC) concentrations as low as 150 cm-3, although 300–500 cm-3 were
more typical. Corresponding non-sea salt fine-mode particle concentrations in
these phases were 1 to 3 µg m-3. Coarse sea salt was observed
at 4–8 µg m-3. While CALIPSO data during the cruise is unavailable,
we suspect that given the regional veering wind shear, highest particle concentrations
were in the MBL. This is supported by NAAPS model data, as well as climatological
analyses and analysis of CALIOP data from immediately after the cruise period.
In between TCs, two significant aerosol injection events were observed,
each lasting ∼ 2.5 days. The first of these increased CPC particle concentrations
to ∼ 1000 cm-3, and average non-sea salt fine-mode particle concentrations
to ∼ 8 µg m-3. We surmise that long-range transport of particles
reduction of convection to allow long-range transport for this case was induced by
a dry-air intrusion between 800 and 600 hPa (∼ 2–4 km) from the Indian Ocean.
This event is perhaps related to backside MJO subsidence and drying. The aerosol
source of this event was likely southwestern Borneo or with some influence of
southern Sumatra. A second more significant event, with CPC counts as high as
5000 cm-3, occurred in the last days of the cruise when an area of very
clear sky formed between two Category 4 TCs. In this case, significant upper-level
subsidence brought dry air down to below 500 hPa (6 km). High winds in the final
stages of the TC inflow arm leading up to this event may have had a role in its
far reaching nature. This air mass was likely dominated by smoke ejection from
southern through southeastern Kalimantan/Borneo, and perhaps the Sulu Sea.
Veering vertical wind shear resulted in aerosol transport largely in the MBL.
While aerosol particle and gas chemistry are subjects of follow-on papers,
there are clear biomass burning signals in both events, particularly in regard to
K+, CO, benzene and methyl iodide in the second event. However, in general,
air chemistry appears to be a mix of industrial pollution and biomass burning,
with sulfur being the most significant element. Black carbon and organic carbon
ranged from 2 % for the cleanest periods, 5–7 % for the aerosol events,
and up to 12 % in Manila bay. Organic carbon was ∼ 30 %, increasing to over 50 % for the cleanest periods.
PCASP derived particle size distributions for more polluted cases was typical
for a mix of pollution and biomass burning, with volume median diameters on the
order of 0.27–0.30 µm. While the PCASP was inoperable for the cleanest
periods, more background conditions in the early part of the cruise showed smaller VMDs, ∼ 0.21 µm.
Frequent rapid decreases in particle concentration and temperature, with
corresponding sharp perturbations in winds, were associated with cold pool events.
Over 20 such cold pool events were observed during the cruise. We noted,
however, that convection in the SCS/ES region is often associated with narrow squall
lines propagating in the monsoonal flow. In the most significant case, convection
was spawned by a severe thunderstorm over Ho Chi Min City, whose cold pool propagated
southward. Once it reached the southwesterly monsoon, another set of convection was
spawned, creating its own northeastward propagating event. Over the next 24 h, multiple sets of convection repeated the cycle, leading to arc cloud formations
extending 100–200 km in latitude propagating across the SCS/ES. Upon reaching the
Vasco, a 1 min-long high wind event (with up to 25 m s-1 instantaneous
winds) coincided with a precipitous fall in fine-mode particle concentrations and
simultaneous spike in coarse-mode sea salt. Satellite and measured recovery times
suggested a 150 km swath was cut through the marine boundary layer by this event.
While cells up to 20 km high are noted, much of the squall line is made up of nonfreezing
clouds with tops of 6 km. Even a cursory view of regional satellite data shows these
squall lines occur frequently in the southwest monsoonal flow. While only tens of km
wide, they can extend 500 km long across the monsoonal flow, likely supported by
low-level veering winds. These events likely cut swaths of aerosol particles out
of the MBL and thus are likely a major driver of regional aerosol life cycle. The
observation of a cold pool well ahead of the convection must be considered in aerosol–convection interaction studies.
Based on the above observations, we discussed implications for aerosol,
cloud, and precipitation interaction studies. While aerosol particles are
clearly identified by the scientific community as having a critical role in
cloud systems, the covariance between the presence of aerosol particles and
the atmospheric boundary layer state creates an intertwined chicken and egg
problem. The potential for confounding studies is significant. Aerosol
injections into the SCS/ES and Sulu Sea regions were clearly modulated by
MJO and TC phenomenon. Dry layers originating in the Indian Ocean influenced
convection thousands of kilometers away. Such features have to be accounted
for in any analysis. However, the significant cloud cover in the region
makes data assimilation for key variables such as water vapor highly
problematic. Aerosol observations also demonstrate substantial clear-sky
bias. Higher resolution scales, such as for convection, impart important
fine features and process that are not easily replicated in models.
Ultimately, this investigation highlights how future studies need tight
constraints on the overall meteorology, including high-frequency phenomena
such as island ejection of smoke by the sea breeze and cold pools.