Introduction
It is recognized that the surface ocean alters the properties of the lower
atmosphere, and so atmospheric albedo and climate (McCoy et al., 2015;
Seinfeld et al., 2016), via the direct and indirect effects of aerosols
(O'Dowd and de Leeuw, 2007). Aerosols are precursors of clouds, which play a
major role in the scattering and absorption of incident solar radiation
(Carslaw et al., 2013), but the concentration, number and chemical
properties of aerosols that act as cloud condensation nuclei (CCN) can also
influence cloud droplet size and number and consequently precipitation and
cloud albedo (Twomey, 1977). Indeed, cloud formation and properties are
sensitive to relatively minor changes in aerosol concentration, particularly
in remote regions (Carslaw et al., 2013). This is particularly the case in
the Southern Ocean, where natural aerosol sources dominate and where CCN
concentrations can range from tens per cubic centimeter in winter to hundreds per cubic centimeter in summer (Andreae and Rosenfeld, 2008), leading to seasonally
variant trends in cloud albedo. However, the relationship between clouds and
aerosols derived from natural sources is poorly understood, and represents a
major uncertainty in the representation of low-level marine clouds and
feedbacks in climate models (Wang et al., 2013; Stephens, 2005). Current
models underestimate cloud over the Southern Ocean, particularly south of
55∘ S, resulting in excess surface shortwave radiation and a warm bias
(Trenberth and Fasullo, 2010; Kay et al., 2016). This discrepancy is
potentially attributable to a variety of factors, chief among which is the
limited understanding of aerosol–cloud interaction and cloud water phase,
compounded by a lack of regional observations and data to advance satellite
retrievals and climate model simulations.
Breaking waves and associated bubble formation are a major source of primary
marine aerosol (PMA), supplying most of the aerosol mass in the marine boundary
layer (MBL) over the remote ocean (Andreae and Rosenfeld, 2008) and
particularly in regions that experience high winds and breaking waves (de
Leeuw et al., 2014). This is reflected in PMA contributing only
∼ 10–20 % of CCN number concentrations over the remote
Pacific Ocean (Blot et al., 2013; Clarke et al., 2013) but up to 55 % over
the Southern Ocean (McCoy et al., 2015). Although PMA is generally regarded
as primarily composed of sea salt, recent reassessments suggest it is highly
enriched in organic matter relative to bulk seawater. Organic material may
in fact dominate submicron aerosol mass (Facchini et al., 2008; O'Dowd et
al., 2004), with the primary organic aerosol (POA) being of biogenic origin
and including bacteria, carbohydrates, polymers and gels (Facchini et al.,
2008; Russell et al., 2010). Although the contribution of POA to the MBL is
uncertain, it may be significant over biologically active oceanic regions,
as suggested by correlations between organic aerosol content and surface
chlorophyll a (Chl a) (O'Dowd et al., 2004). There is also similarity in the
composition of aerosol and surface ocean organics, and organically enriched
submicron particles have been produced experimentally using surface
seawater conditions (Quinn and Bates, 2011). Indeed, the degree of organic
enrichment may influence both the type and size of aerosols, as well as
properties such as aerosol light scattering and water uptake (Vaishya et
al., 2012).
It is well-established that biologically productive regions are
characterized by elevated concentrations and emissions of a range of
compounds that may influence aerosol production, composition and properties
(Meskhidze and Nenes, 2010; Gantt and Meskhidze, 2013; de Leeuw et al.,
2014). However, the oceanic influence on atmospheric composition is not only
attributable to PMAs but also to secondary marine aerosols (SMAs) that are
produced during gas-phase reactions of volatile organic compounds (VOCs).
Although SMAs have less impact upon aerosol mass, they potentially have a
large influence on aerosol number (Meskhidze et al., 2011). The
biogeochemical origin of SMAs is reflected in their seasonality, with Aitken
and accumulation mode aerosol number concentrations dominated by secondary
particles in summertime (Clarke et al., 2013; Cravigan et al., 2015). Research
into SMAs has primarily focussed on dimethylsulfide (DMS), the primary
natural marine source of volatile sulfur, in response to early hypotheses
related to its potential role in climate feedback processes (Charlson et
al., 1987). The CLAW hypothesis linked the production of the DMS precursor,
dimethylsulfoniopropionate (DMSP), by phytoplankton and subsequent DMS
emission and oxidation to sulfate aerosol, to CCN formation and changes in
cloud cover. Although well-studied, this hypothesis remains unproven and
there is a lack of consensus, with a recent review identifying uncertainties
regarding the role of DMS in aerosol production in the MBL (Quinn and Bates,
2011). However, there is evidence that DMS may play a role in cloud
formation over larger spatial and temporal scales, via entrainment from the
free troposphere (Carslaw et al., 2010).
The fundamental tenet of the CLAW hypothesis, of feedback between surface
ocean biogeochemistry and climate, may be applicable via a broader spectrum
of precursor species. Recent research has shown increasing complexity of
potential aerosol source pathways, involving a variety of chemical species,
processes and interactions (Vaattovaara et al., 2006). In addition to DMS, a
variety of other gaseous aerosol precursors that originate from
phytoplankton, bacterial and photochemical sources at the sea surface may
undergo physical and chemical transformation to produce new particles in the
MBL (Ciuraru et al., 2015). These SMA precursors include volatile organic
species, such as carboxylic acids, isoprene, monoterpenes, halocarbons,
iodine oxides and iodine (Vaattovaara et al., 2006; Sellegri et al., 2005).
A biological source of these SMAs has been inferred from the spatial and
temporal correlation between phytoplankton blooms and cloud microphysics
(Meskhidze et al., 2009; Meskhidze and Nenes, 2010; Lana et al., 2012). The
presence and concentration of SMA precursors in the MBL may be dependent
upon plankton abundance and community composition, and consequently their
influence on aerosol formation will show spatial and seasonal variability
(O'Dowd et al., 2004).
New particle formation may be suppressed by the interaction of aerosol
precursors and SMAs with preexisting aerosol, for example, by absorption of
ammonia and gaseous sulfuric acid by coarse-mode sea salt aerosol (SSA;
Cainey and Harvey, 2002). Conversely, existing particles may grow via
condensation, which enhances their CCN capacity (Clarke et al., 2013). It has
also been proposed that organic acids combine with sulfuric acid to create
the critical nucleus required for aerosol formation (Zhang, 2010; Almeida
et al., 2013). However, nucleation events over the open ocean remain elusive
(O'Dowd et al., 2010; Chang et al., 2011; Willis et al., 2016), making it
difficult to elucidate the primary pathways and reactants, and consequently
they are currently regarded as of low significance to marine aerosol
formation. Following nucleation, the aerosol distribution is modified by
aerosol–aerosol interaction, heterogenous reactions and removal processes,
including coagulation and condensation, resulting in the longest-lived
aerosol component being in the accumulation mode (0.06–0.4 µm). With
such a wide variety of potential precursors and inorganic–organic
interactions affecting nucleation and CCN activation, the modeling of
aerosols and their indirect influence on cloud radiative properties over the
remote ocean presents a major challenge (Seinfeld et al., 2016).
The production and transfer of aerosol precursors from the ocean surface is
also dependent upon physical factors. Exchange across the air–sea interface
is primarily controlled by near-surface turbulence, which is dependent on
wind and waves. For practical purposes, this is represented by a kinetic
factor, the transfer velocity k, which is generated with wind speed
parameterizations (Nightingale et al., 2000; Ho et al., 2006). Although
wind speed provides a reasonable broad-scale proxy for kinetic transfer,
other factors such as fetch, wave development, wind–wave direction and
surfactants, also influence k and so generate variation in gas exchange and
deviation from k–wind-speed relationships. For example, most k–wind-speed
parameterizations do not explicitly capture the solubility effects
associated with bubbles (Blomquist et al., 2006), although the COAREG gas
transfer model incorporates this factor into a physically based flux
algorithm (Fairall et al., 2003, 2011). Biogeochemical
gradients near or at the ocean surface are also not considered, despite
their potential to alter the air–sea exchange of gases, PMAs and SMAs
(Facchini et al., 2008; Calleja et al., 2013).
Previous related research campaigns have examined the biogeochemical and
physical factors influencing oceanic DMS and CO2 fluxes, as summarized
in Supplement Table S1, but few have linked this to the physical controls of
air–sea exchange and variation in the aerosol and trace gas composition of the
MBL. Similarly, other campaigns with an atmospheric focus, such as MAP (Marine Aerosol Production; Decesari et al., 2011), have carried out detailed studies of aerosol
chemistry but have not interpreted this with regard to surface ocean
biogeochemistry. To address this, the Surface Ocean Aerosol Production
(SOAP) campaign was initiated, with the primary aim of characterizing the
variation in aerosol composition and concomitant marine sources, processes
and pathways in the southwest Pacific. SOAP utilized a multidisciplinary
framework, encompassing surface ocean biology and biogeochemistry, transport
and air–sea exchange with a characterization of aerosol number and
composition, to establish controls on aerosols and gas exchange. The
campaign consisted of two voyages – a pilot study, PreSOAP, which carried
out a regional survey and established sampling strategies, and the
following SOAP voyage – in biologically productive frontal waters along the
Chatham Rise, east of New Zealand (see Fig. 1). Building upon the
approaches used in previous studies, the SOAP campaign targeted three
phytoplankton blooms of differing plankton community composition to
determine their respective influences on biogeochemistry, gas exchange and
MBL composition. This paper details the regional context, sampling
strategy, environmental conditions and some preliminary results for the SOAP
campaign.
(a) An ocean color image (10/2/11) during the PreSOAP voyage,
showing phytoplankton blooms on the western Chatham Rise region along
44∘ S (data courtesy of NASA). (b) The SOAP voyage track in the Chatham
Rise region, overlain by sea surface temperature (∘C), with the study
region (box) indicated in the inset bathymetric map of New Zealand.
Regional context
The southwest Pacific has many features in common with the Southern Ocean,
as it is characterized by low anthropogenic and terrestrial aerosol loading,
long ocean fetch and high wind speed, making it an optimal location for
examining the marine contribution to aerosol production. One of its more
biologically productive regions lies east of New Zealand, where the
subtropical front (STF) extends as a tongue of elevated phytoplankton
production (Murphy et al., 2001) along 43.0–43.5∘ S over the Chatham
Rise (see Fig. 1a). This arises from the confluence of warmer saline
subtropical waters that are relatively depleted in macronutrients, with
fresher cooler subantarctic waters containing elevated macronutrients but
being depleted in iron (see Fig. 1b; Boyd et al., 1999). Mixing across the front
alleviates nutrient stress, which, combined with a relatively stable water
column, promotes primary production (Chiswell et al., 2013). Ocean color
climatologies show a monthly mean Chl a of 0.6 mg m-3, reaching
∼ 1 mg m-3 over the Chatham Rise in spring (Murphy et
al., 2001), and the region is characterized by elevated marine particle
export, secondary production and fish stocks (Nodder et al., 2007;
Bradford-Grieve et al., 1999). In spring the phytoplankton community
composition varies with water mass, with diatoms dominating the STF,
cryptophytes, prasinophytes and dinoflagellates being more prevalent in
subtropical waters, and photosynthetic nanoflagellates dominating
subantarctic waters (Chang and Gall, 1998; Delizo et al., 2007). The STF
also supports spatially extensive coccolithophore blooms (Sadeghi et al.,
2012) and is situated on the northern edge of the “Great Calcite Belt”
(Balch et al., 2011), a latitudinal band of elevated backscatter attributed
to coccolithophore liths. Surface mixed layer nutrients vary spatially in
response to mixing of the water masses and seasonally due to phytoplankton
uptake, with the evolution of nutrient stoichiometry and grazing determining
the succession and duration of different phytoplankton blooms (Chang and
Gall, 1998; Delizo et al., 2007). The STF is characterized by significant
gradients in pCO2 associated with phytoplankton blooms, with current
global climatologies indicating the region east of New Zealand as a
significant carbon sink (> 1 mol C m-2 yr-1; Landschuetzer et al., 2014).
The waters south of New Zealand are characterized by high wind speeds, which
drive the disproportionate contribution of this region to global ocean
CO2 uptake. Here, wind, waves and currents develop unhindered by
land, and strong persistent westerlies act over long fetch to generate large
swells that propagate northeast influencing the wave climate off New
Zealand. While this wave energy is attenuated closer to land in the eastern
Chatham Rise, the average wave energy is still 75 % of values south of New
Zealand, where annual mean wave heights exceed 4 m. Subantarctic waters south
of the Chatham Rise region provided a prime location for a dual tracer
release experiment (SAGE; Harvey et al., 2011), aimed at constraining k at
high wind speeds. Comparison of the SAGE k–wind-speed parameterization with
those generated in other regions and using different techniques showed
generally good agreement (Ho et al., 2006); this may be interpreted as
indicating that regional influences on exchange may be less important,
supporting the application of a universal wind speed parameterization.
Nevertheless, other factors, such as wave age, duration and height do
influence gas exchange in this region (Smith et al., 2011; Young et al.,
2012). The elevated winds also influence the transfer of aerosols and
precursors, as reflected by a zonal band of elevated sea spray aerosol mass
and water-insoluble organic matter over the Chatham Rise region (Vignati et
al., 2010).
Both models and measurements indicate that DMS is a significant contributor
to total non-sea-salt sulfate (nssSO4) in the Southern Hemisphere (Gondwe et
al., 2003; Korhonen et al., 2008). However, a paucity of observational data
in the Southern Ocean has hindered the development of global climatologies for
surface seawater DMS (DMSsw), with the region southeast of New Zealand
represented by only a few data points in a recent DMS climatology (Lana et
al., 2011). Despite this shortcoming, this climatology provides a realistic
representation of atmospheric DMS and total sulfate when applied in
aerosol–climate global climate models, particularly over the Southern Ocean
(Mahajan et al., 2015). Seasonal variability in atmospheric DMS is apparent
at stations around New Zealand and south of 44∘ S (Blake et al., 1999), with
concentrations of 100–200 pptv and maximal values associated with the
transport of DMS from waters to the south in summer (Harvey et al., 1993; de
Bruyn et al., 2002; Wylie and de Mora, 1996). Corresponding seasonality in
nssSO4 was observed, with a maximum (0.8–1.5 µg m-3) in early
austral summer at the start of the year, decreasing in late summer to
0.1–0.4 µg m-3 through autumn and winter (see Fig. 2;
Sievering et al., 2004; Allen et al., 1997). For comparison, coarse SSA
dominates the aerosol mass at Baring Head, with concentrations of 6–10 µg m-3 (Jaeglé et al., 2011; Spada et al., 2015). Similar
seasonal cycles of DMS and nssSO4 were recorded at Cape Grim (Ayers, 1991),
and the observed diurnal inverse correlation between sulfur dioxide and DMS
at Baring Head was applied to estimate yield and the potential contribution to
aerosols (de Bruyn et al., 2002). Consistent seasonal trends between
activated particles and cloud droplet number concentration were also
apparent, with a summer maximum over the Southern Hemisphere (Boers et al.,
1996, 1998), related to phytoplankton production (Thomas et al., 2010).
Overall, the temporal trends in aerosol precursors and pathways do not
follow that of wind speed and other physical drivers but instead reflect
biological processes inferring control by surface ocean biogeochemistry
(Korhonen et al., 2008).
Non-sea-salt sulfate concentrations plotted against day of year at
different New Zealand coastal atmospheric monitoring sites.
SOAP work programmes and observations
A number of parameters were measured (see Table 2) in three interlinked work
programmes during the SOAP voyage, as indicated in Fig. 8 and detailed
below.
Parameters sampled during the SOAP voyage. Key: C – continuous; D
– discrete; W – workboat; *indicates instrument sampling on common aerosol
inlet.
Measurement
Mode
Instrument
WP1 Atmospheric
Organic nuclei production
C*
Ultrafine organic tandem differential mobility analyzer (UFO-TDMA)
Aerosol water uptake and volatility
C*
Volatility humidity differential mobility analyzer (VH-TDMA)
Nucleation/Aitken mode size spectra
C*
Scanning mobility particle sizer (SMPS)
Condensation nuclei counts
C*
Condensation particle counter (CPC)
Accumulation mode aerosol number
C
Passive cavity aerosol spectrometer probe (PCASP)
Cloud condensation nuclei
C*
CCN spectrometer
Aerosol filter chemistry – major ions
C
Hi-vol, cascade, ion chromatograph
Black carbon
C*
Aetholometer
PM10 aerosol filters
C
Organic functional groups by FTIR and inorganic composition by ion beam analysis
Column aerosol
D
Sun photometer (Microtops II)
Nascent sea spray composition via bubble burst of seawater samples
D
Chamber experiments
DMS
C
MesoCIMS (chemical ionization mass spectrometry)
CO2 and methane
C
Picarro CRDS
Halocarbons, iodine and halogen oxides
C
μ-Dirac electron capture detector–gas chromatograph and multi-axis differential optical absorption spectroscopy (Max-DOAS)
VOCs (acetone, DMS, acetonitrile, methanol, methanethiol, isoprene, monoterpenes, acetaldehyde)
C
Proton transfer reaction mass spectrometer (PTR-MS)
VOCs C5 to C15
D
Pre-concentration and TD-GC-FID/MS
Aldehydes, ketones (incl. dicarbonyls), C2 to C8
D
Derivatization and HPLC
WP2 physics
DMS flux
C
MesoCIMS (chemical ionization mass spectrometry)
CO2 EC flux
C
LI-COR infrared gas analyser (IRGA), sonic anemometer motion sensors
DMS gradient flux
D
Catamaran, SCD-GC
Near-surface T and S
D
Conductivity–temperature–depth (CTD)
Near-surface stratification
C
Spar buoy – temperature array, microcats
Near-surface turbulence
C
Vector, FastCat
Sea state
C
NOAA Wavewatch III
Whitecap coverage
D
Campbell Scientific 5-megapixel Camera
Meteorological conditions
C
Automatic weather station (AWS)
Bulk fluxes
C
Eppley radiometers, rain gage; Eppley Precision Spectral Pyranometer (PSP)
MBL height and stability
D
Radiosonde
WP3 ocean biogeochemistry
Chlorophyll a
C, D, W
Ecotriplet
Backscatter and β660 backscatter
C
Ecotriplet
pCO2
C
IRGA
pH
C, D, W
Spectrophotometer
Dissolved inorganic carbon (DIC)
D
Nutrients
D, W
Colorimetric autoanalyzer
Dissolved organic carbon (DOC)
D, W
High-temperature catalytic oxidation (HTCO)
CDOM
D, W
Spectrophotometer
Particulate organic and total carbon and nitrogen and isotopes (POC/PON/PC/PN/13C/15N)
D
Mass spectrometer
Fatty acids and alkanes
D, W
Dissolved DMS
C,
MiniCIMS (chemical ionization mass spectrometry)
Dissolved DMS
D, W
SCD/FPD (flame photometric detector)
DMSP and processes
D, W
SCD
Pigments
D
HPLC
Microbial community abundance
D, W
Flow cytometry
Phytoplankton identification/counts
D, W
Optical microscopy
Microzooplankton
D, W
Optical microscopy
The distribution and composition of aerosols, precursors and trace
gases in the MBL
Conceptual figure of the parameters, processes and vertical range
measured during SOAP, with the integrated work programmes (WP) indicated on
the left of the figure. CN: condensation nuclei. NSS: non sea salt.
Aerosol number concentration, size distribution, composition, water uptake
and CCN concentration were measured semicontinuously during SOAP to address
the overall paucity of aerosol observations and the apparent rarity of
nucleation events over the remote ocean. These were characterized by a
suite of instruments covering a particle size range of 0.01 to 10 µm
(see Fig. 9 and Table 2), which enabled the determination of the
size-dependent contribution of PMA and nssSO4 to aerosol and CCN
concentrations. Aerosol characterization identified variable Aitken and
consistent submicron-sized accumulation and coarse modes, with the
submicron aerosol mass dominated by secondary aerosol with ammonium
sulfate/bisulfate under light winds and with an increase in sea salt
proportion as local winds increased. Ongoing data analysis is examining
whether significant nucleation events occurred.
Aerosol characterization during SOAP, indicating size spectral
(red) and total count (black) range for each instrument, relative to
aerosol size and mode. Ambient RH measurement was used for RH correction of
the PCASP, Hi Vol and SMPS, and diffusion driers (Silica Gel) were used on
the inlet of the UFO-TDMA and VH-TDMA.
The operational mode for underway aerosol measurement was to slowly steam at
1–2 kn into the prevailing wind, across an area of high biological
productivity or a significant air–sea gas gradient, generally between noon and
14:00 when solar irradiance was maximal. The common aerosol inlet developed
during PreSOAP allowed uncontaminated air from above the bridge to be
sampled when the wind was on the bow, thus minimizing interference from ship
stack emissions. Contamination events were screened out using a real-time
clean-sector sampling “baseline” flag and switch (Harvey et al., 2017), enabling the clean collection of integrated samples. Although the
vessel exhaust was the primary contaminant, other potential sources included
the workboat and recirculation of polluted air around the ship, and longer-range terrestrial influences were also assessed. Measurements of black
carbon using an aethalometer and CO2 by high-precision cavity
ring-down laser spectroscopy (CRDS) provided two independent variables for
detecting contamination events, and some VOCs, measured by proton transfer reaction mass spectrometer (PTR-MS; see Table 2), were also used as indicators of diesel combustion. The vessel was
orientated into the wind as often as possible, which resulted in a high
frequency (∼ 75 %) of baseline sector conditions during the
SOAP voyage. Clean marine air periods were defined post-voyage, using the
baseline wind sector (225–135∘ relative to bow and wind speed
greater than 3 m s-1), black carbon concentrations (less than 50 ng m-3) and back trajectories, and indicated minimal terrestrial impact
(periods when the minimum number of hours over land in 72 h back
trajectory is zero), with periods of workboat operations removed. An
ensemble of Hybrid Single-Particle Lagrangian Integrated Trajectory
(HYSPLIT) model back trajectories (Draxler and Rolph, 2013) was run for each
hour of the voyage, and NAME back trajectories were calculated for every 3 h (Fig. 5; Jones et al., 2007). Figure 10 shows particle number and
CCN concentrations compared to the number of hours the 72 h back
trajectory spent over land calculated from HYSPLIT trajectories. Particle
concentrations were generally higher during periods of terrestrial influence
(see DoY 52 and 60; Fig. 10), with average particle number concentrations
of 1122 ± 1482 cm-3, double that observed for clean marine air.
Ion beam analysis also revealed the presence of silicate and aluminium on
ambient submicron filter samples, suggesting a terrestrial source and supporting
the back-trajectory modeling of continental outflow.
(a) Marine boundary layer CN concentrations (cm-3;
top; CPC3772 in blue, CPC3010 in red), (b) CCN concentrations (middle; cm-3)
and (c) number of hours over land indicated by 72 h back trajectory
(bottom; 27-member ensemble average). Bloom occupation periods are indicated
by the vertical shaded bars and bloom labels at the top of the figure.
During the initial occupation of B1 under light winds, the particulate
matter (PM10) total ion mass was 9.5 µg m-3 compared to
subsequent samples under higher winds in the range 20–50 µg m-3. The dominant components of the inorganic mass fraction were
sea salt ions and nssSO4, although a measurable organic fraction was also
present (see below). The NaCl mass in light winds during B1 was 6.6 µg m-3 with > 95 % being > 3 µm in diameter,
relative to 32.5 µg m-3 under stronger winds during B3b.
Although 72 % was > 3 µm, the largest difference in mass
occurred in the 1.5 to 3 µm size range. In contrast, the mass of
nssSO4 was predominantly submicron sized; B1 exhibited the largest nssSO4
mass at 2.0 µg m-3 with 85 % in sizes < 1 µm, whereas in B3b, the nssSO4 mass was much lower at 0.6 µg m-3
with 76 % with sizes < 1 µm. These results
confirm the influence of both physical and biogeochemical processes on
aerosol composition.
Voyage particle number concentrations during clean marine periods averaged
534 ± 338 cm-3, with CCN concentrations of 178 ± 87 cm-3 (±1 sd) at 0.5 % supersaturation and an average particle
fraction activated into CCN of 0.4 ± 0.2. Bloom average particle
number concentrations ranged from a minimum of 385 ± 96 cm-3 in
B3b to a maximum 830 ± 255 cm-3 at the start of B2 (Fig. 10).
B1 displayed the highest CCN activation ratio, of 0.5 ± 0.2,
potentially due to the combination of low wind speeds, large biogeochemical signals
and SMA fluxes. Comparison of the inorganic ion mass, determined from
high-volume sampler filters, between the different blooms does not support
the conclusion that the B1 activation ratio was higher simply because
particles were larger. As the median particle diameters during clean marine
periods were consistent between the three blooms, this suggests that
particle composition, secondary organics or coagulation may have impacted
CCN activation at B1. These findings are supported by preliminary results
from an application of the ACCESS-UKCA model (M. Woodhouse, personal communication, 2017), which
simulated the additional impact of emissions of marine secondary organic
carbon under the conditions determined during SOAP. In contrast, the average
CCN activation ratio for B3a was lower at 0.13 ± 0.06. Nucleation mode
particles (10 and 15 nm) were measured by ultrafine organic tandem
differential mobility analyzer (UFO-TDMA; Vaattovaara et al., 2005) and
Aitken mode particles (50 nm) by UFO-TDMA and a volatility and
hygroscopicity tandem differential mobility analyzer (VH-TDMA; Johnson et
al., 2004; Villani et al., 2008). This analysis typically identified a
significant (up to 50 % volume fraction) secondary organic component
during sunny conditions in bloom regions, particularly during B1. The TDMA
results provided further evidence for secondary organic aerosol processing
of the dominant secondary nssSO4 mode during B1. Deliquescence measurements
(VH-TDMA) indicate that the Aitken mode population is largely comprised of
neutralized nssSO4, i.e., ammonium sulfate. Small and sporadic
contributions to the Aitken mode from a nonhygroscopic component (number fraction up
to 0.4) and a highly hygroscopic component (number fraction up to 0.3) were
observed in addition to the secondary nssSO4 mode (number fraction of
0.6–1). The water uptake and volatility of the sporadic highly hygroscopic
mode indicates that this may be composed of PMA.
The in situ aerosol size, number and composition measurements in the MBL were
complemented by in vitro chamber measurements of nascent SSA to determine the PMA
organic volume fraction and water uptake properties. Nascent SSA filter
samples were analyzed using Fourier transform infrared spectroscopy (FTIR)
for organic functional groups (Russell et al., 2011) and ion beam analysis
for inorganic concentrations (Cohen et al., 2004). Measurements of the
hygroscopic growth factor and the volatile fraction up to 450 ∘C
for 50–150 nm particles using the VH-TDMA were compared with those of
reference inorganic samples (e.g., sea salt, ammonium sulfate) to determine
their organic volume fractions (Modini et al., 2010). Complementing the
VH-TDMA, the UFO-TDMA provided further information on the organic content of
particles of 50 nm and down to 10 nm. The bubble chamber observations
indicated that the PMA contained a substantial primary organic fraction.
VH-TDMA results indicate that the Aitken mode PMA was primarily nonvolatile
(78–93 %), with an average organic volume fraction of 51 % (ranging from
39 to 68 %), and the UFO-TDMA results show an organic volume fraction (OVF) ranging from 35 to 45 %.
These results are consistent with observations in the North Pacific and
Atlantic, for which an Aitken mode volatile fraction of the order of 15 %
and OVF of 0.4–0.8 have been observed (Quinn et al., 2014). FTIR analysis
indicated that the POA aerosol in the chamber experiments was largely
composed of hydroxyl functional groups, with minor contributions from
alkanes, amines and carboxylic acid groups, consistent with previous observations (Russell et al., 2011).
Although DMS was a primary focus of measurements during SOAP, a wide variety
of other VOCs that potentially contribute to secondary organic aerosol
formation were also measured. Halogens and halogen oxides were measured
using multi-axis differential optical absorption spectroscopy (Max-DOAS) and
electron capture detector–gas chromatography (ECD-GC). Iodine has been
identified as a potentially important precursor of nucleation in coastal
regions (Sellegri et al., 2005), and SOAP provided an opportunity to relate
the presence of halogen oxides to phytoplankton biomass and composition in
the surface ocean and nucleation events in the MBL. A high-sensitivity PTR-MS carried out measurements continuously in
H3O+ mode in the range of m/z 21–m/z 155 throughout the voyage
(Lawson et al., 2017). Aldehydes, ketones and dicarbonyls were
measured using 2,4-dinitrophenylhydrazine (2,4-DNPH) cartridges and high-performance liquid chromatography (HPLC; Lawson et al., 2015), and a range
of VOCs were sampled using adsorbent tubes and later analyzed via thermal
desorption–gas chromatography–flame ionization detection–mass
spectrometry (TD-GC-FID/MS). These measurements identified a
positive relationship between DMS (m/z 63), acetone (m/z 59) and
methanethiol (m/z 49), indicating common biological drivers (Lawson et al.,
2017).
The first in situ measurements of aqueous phase SMA precursors dicarbonyls,
glyoxal and methylglyoxal were obtained over the remote Southern Ocean
during SOAP (Lawson et al., 2015). Parallel measurements of known dicarbonyl
precursors, measured by PTR-MS, were used to calculate the expected yields
of glyoxal and methyl glyoxal, which accounted for < 30 % of
observed mixing ratios indicating an unidentified source of dicarbonyls
(Lawson et al., 2015). This was corroborated by inclusion of SOAP glyoxal
measurements obtained by Max-DOAS measurement in a global database, which
concluded that the missing glyoxal source was an order of magnitude greater
than identified sources (Mahajan et al., 2014). Surface mixing ratios of
glyoxal converted to vertical columns, were significantly lower than average
vertical column densities (VCDs) from satellite retrievals, possibly
reflecting the difficulty of retrieving low glyoxal VCDs over the ocean or,
alternatively, incorrect assumptions about the vertical distribution of glyoxal in the
atmosphere (Lawson et al., 2015).
Rates and controls of volatile and precursor emissions at the air–sea
interface
DMS measurements were made using three different instruments during SOAP
(see Table 2); an atmospheric pressure ionization–chemical ionization mass
spectrometer (API-CIMS) continuously monitored DMS in both phases (Bell et
al., 2015), a PTR-MS monitored DMSa (Lawson et al., 2017), and
discrete water measurements were made using a sulfur chemiluminescence
detector gas chromatograph (SCD-GC; Walker et al., 2016). Intercomparison of
sulfur measurements is not easily or routinely performed (Bell et al.,
2012), particularly at sea. Seawater DMS measurements (CIMS and SCD-GC)
compared well during SOAP (Walker et al., 2016), and the SCD-GC technique
also compared well with traditional gas chromatography (with
flame photometric detector) in an international intercalibration exercise
(Swan et al., 2014). Intercomparison of the PTR-MS and SCD during SOAP
involved analysis of two air samples and two diluted DMS gas standards with
a concentration range of 158–354 ppt. The instruments showed very good
agreement, with a mean difference of 5 % and a maximum of 10 %.
Although the majority of DMS flux estimates to date have been derived by
applying an independently determined transfer velocity (k) to the measured
DMS gradient at the ocean surface (ΔDMS), there has been a recent
increase in direct micrometeorological measurements of DMS flux.
Measurements at 10–30 min resolution show considerable variability in
flux, which may reflect methodological artefacts or inherent variability in
the distribution of DMS. SOAP provided a platform for comparing EC flux measurements of DMS using API-CIMS (Bell et al., 2015),
with a gradient flux technique using a drogued catamaran within 1 km of the vessel (Smith et al., 2017). The gradient flux
technique is less direct than EC but provides an alternative reference on a
platform that is relatively free of shipboard airflow distortion. The EC
system sampled from an intake on the ships bow, with flux instruments
mounted on the foremast 12.6 m above sea level and the air pumped to a
containerized laboratory on the foredeck. Additional meteorological
measurements were obtained from a weather station above the bridge. Both
sites are subject to airflow distortion which is azimuthally dependent
(Popinet et al., 2004). The catamaran sampling framework, which consisted of
four air intakes distributed vertically on a 5.3 m mast, sampled closer to
the water surface where gas gradients are largest. Flux measurements were
augmented by continuous near-surface measurement of physical parameters
using a range of sensors attached to a spar buoy, with stratification
determined by temperature sensors at 0.5 m intervals (Walker et al., 2016) and turbulence determined by a vector acoustic doppler velocimeter at 0.6 m
depth. This permitted comparison of kDMS estimates with near-surface
upper-ocean turbulence at a distance from the vessel (Smith et al., 2017). Wave-breaking whitecap coverage was monitored using a Campbell
Scientific 5-megapixel camera (cc5mpx) located on the starboard side of the
vessel (Scanlon and Ward, 2016). This provided an indicator of bubble
entrainment, which contributes to the differential transfer rate of DMS and
CO2 due to their different solubilities (Blomquist et al., 2006; Bell
et al., 2017).
Although SOAP primarily focussed on DMS fluxes, EC measurements of CO2
flux were an important adjunct measurement for providing insight into gas
exchange mechanisms and controls and improving gas transfer algorithms for
gases of differing solubilities. Four LI-COR infrared gas analyzers were used
for eddy covariance flux measurements of CO2 during SOAP, following the
initial trials on PreSOAP. Comparison of EC measurements with wet and dry
incoming gas streams and an empirically based post-processing correction indicated that only gas stream drying produced robust CO2 flux and
kCO2 estimates (Landwehr et al., 2014). A detailed examination of ship
motion and airflow distortion effects resulted in a significant reduction in
the scatter in the CO2 eddy covariance data (Landwehr et al., 2017).
The EC-derived kCO2 estimates provided a better correlation with a
linear fit to the EC friction velocity than with the 10 m neutral wind
speed (u10N) and showed good agreement with dual tracer-derived estimates
from the SAGE experiment conducted in this region in March–April 2004 (Ho et
al., 2006). Measurement of DMS and CO2 fluxes also provided further
constraint of k parameterizations based upon wind speed and the opportunity
to assess the influence of bubbles on gas exchange at high wind speeds. DMS
fluxes derived by EC and gradient flux techniques showed good agreement
(Bell et al., 2015; Smith et al., 2017) and confirmed previous
observations that gas transfer is a linear function of wind speed at low to
intermediate winds (Blomquist et al., 2006; Yang et al., 2011). Despite winds reaching 20 m s-1 during the latter part of SOAP,
insufficient data were obtained to draw conclusions regarding the reported
deviation of kDMS under high winds (Bell et al., 2015). However, SOAP
provided a novel estimate of the size of the EC flux footprint and the
temporal–spatial mismatch between DMSsw and shipboard measured fluxes,
highlighting the importance of considering skew in flux estimates arising
from nonlinear distribution of DMSsw (Bell et al., 2015).
A further objective of SOAP was a comparison of measured DMS fluxes with
calculated estimates from the COAREG model (Fairall et al., 2011) based on
ΔDMS, to assess potential discrepancies with modeled fluxes
(Marandino et al., 2008; Walker et al., 2016). Potential factors examined
here included air and water stability and the influence of the SSM. Despite
the agreement between DMS flux estimates by the two micrometeorological
techniques, there was significant departure from COAREG predictions (Fairall
et al., 2011) on occasions, suggesting the influence of unidentified
processes (Smith et al., 2017). One example was the
suppressed DMS flux during a period of atmospheric stability and reversed
heat flux during B2. Concurrent EC flux measurement for DMS and CO2
also provided an opportunity to assess other influences on k. The DMS
flux data indicate that the kDMS–wind-speed relationship was relatively
insensitive to surface biogeochemistry or wave action during SOAP (Bell et
al., 2015). In addition, SOAP data were used to parameterize whitecap
coverage against wind speed and identify the fact that maturing waves may obscure
and lead to underestimation of the variability of breaking waves (Scanlon and
Ward, 2016).
Surface ocean biogeochemical influences on aerosols and
volatiles
Surface mapping of DMSsw and pCO2, using API-CIMS and IRGA, respectively
(Bell et al., 2015), were critical to the SOAP voyage strategy and the aims
of the two work packages discussed above. These measurements also provided
insight into the covariance of DMS sources and CO2 sinks in surface
waters and establish the importance of this region to global budgets. The
New Zealand Coastal (NEWZ) province (Longhurst, 1998), which includes the frontal region (STF)
studied during SOAP, is characterized in the global DMS climatology by
year-round low DMS concentrations with a maximum < 2 nmol L-1
(Lana et al., 2011). This infers that this region has some of the lowest
global DMSsw concentrations, in marked contrast to the adjacent South
Subtropical Convergence (SSTC) province, which occupies the remainder of the
35–50∘ S latitude band and also accommodates the STF and is characterized
by a mid-summer maximum of 10 nmol L-1 DMS. This discrepancy between
the two regions likely reflects the low number of DMS observations for the
NEWZ province in the climatology (n=6; Lana et al., 2011). Previous DMSsw
measurements in subantarctic waters south of the Chatham Rise, and east of
Tasmania in the SSTC biome (Archer et al., 2011; Griffiths et al., 1999),
are consistent with this climatological estimate, whereas larger unpublished
surveys have recorded elevated surface DMSsw during austral spring (October 2000), with a mean DMSsw of 4.5 (±6.8) nmol L-1 on the Chatham
Rise (BOX voyage, M. Harvey et al., personal communication, 2017). Combining these measurements with data
from the SOAP campaign (mean DMSsw: 6.6 nmol L-1, Bell et al., 2015) gives a
weighted-mean DMSsw of 5.3 nmol L-1 (n=5300, see Table 3),
confirming that DMSsw in the NEWZ province is currently underestimated, and
is in fact more typical of the SSTC province. Although the PreSOAP and SOAP
sampling strategy of focussing on phytoplankton blooms may introduce bias
towards higher DMSsw, the BOX voyage, which had broad spatial coverage of
subtropical and subantarctic waters between 39.5 and 47∘ S, gave a similar
mean DMSsw to the weighted mean for all voyages. The elevated DMSsw was
reflected in the EC flux measurements during SOAP, which recorded maximum
and mean fluxes of 100 and 16.3 µmol S m-2 d-1,
respectively, (Bell et al., 2015), which exceed the climatological
mean of > 10 µmol S m2 d-1 for the SSTC region
(Lana et al., 2011). In addition, the high MBL DMS concentrations of 1000 ppt recorded during SOAP exceed DMSa at coastal stations on the New Zealand
North Island in summer (Harvey et al., 1993; de Bruyn et al., 2002; Wylie
and de Mora, 1996). Although seasonally constrained, the SOAP measurements
provide evidence that regional DMS emissions are significant in this region of the southwest Pacific. The increased dataset of regional concentrations and flux will
allow further refinement of global climatologies, such as the Global Surface
Water DMS Database and the Surface Ocean CO2 Atlas (SOCAT).
DMS data for the SW Pacific region east of New Zealand. SD: standard deviation; n: number of measurements; FPD-GC: flame
photometric detector – gas chromatograph; PFPD – pulsed flame photometric
detector – gas chromatograph; MIMS – membrane inlet mass spectrometer; miniCIMS – atmospheric
pressure chemical ionization mass spectrometer; SCD – sulfur chemiluminescent
detector; Climatol. – climatology; n/a – not available.
Voyage
Date
Latitude
Longitude
Mean DMS
SD
n
Method
Reference
(nmol L-1)
BOX
October 2000
39.5–47∘ S
170∘ E–179∘ E
4.55
6.8
482
FPD-GC
this paper
November 2005
49–50∘ S
175∘ E
1.75
–
2
FPD-GC
Kiene et al. (2007)
SAGE
April 2006
41–46.6∘ S
172.5∘ E–178.5∘ E
1.06
0.9
6
PFPD
Archer et al. (2011)
PreSOAP
February 2011
42.5–44∘ S
174∘ E–178∘ W
2.2
2.0
736
MIMS
this paper
SOAP
March–April 2012
41.7–46.5∘ S
172∘ E–179∘ W
6.36
4.4
4132
miniCIMS
Bell et al. (2015)
SOAP
March–April 2012
41.7–46.5∘ S
172∘ E–179∘ W
11.5
9.2
22
SCD
Walker et al. (2016)
SW Pacific
Weighted mean
39.5–50∘ S
170∘ E–179∘ W
5.6
5380
this paper
NEWZ
35–55∘ S
170∘ E–170∘ W
0.05–2.0
6
Climatol.
Lana et al. (2011)
SSTC
35–50∘ S
170∘ E–170∘ W
0.05–10
n/a
Climatol.
Lana et al. (2011)
The spatial variability of DMSsw was related to surface ocean
biogeochemistry and bloom type by measurement of a suite of ancillary
parameters in underway mode, including temperature and salinity, Chl a, chromophoric
dissolved organic matter (CDOM), β660 backscatter, dissolved
oxygen and pCO2 (see Tables 1 and 2). The vertical variability of
DMSsw and the dissolved and particulate pools of its precursor DMSP were
quantified in the surface mixed layer at stations within each bloom and
related to plankton biomass and community composition, nutrient and organic
composition, and physical drivers (see Supplement Table S2). Process studies of
DMSP cycling included deck incubations examining the bacterially mediated
pathways of DMSP cleavage and demethylation in relation to different bloom
dynamics (Lizotte et al., 2017). DMSP concentrations were relatively
high, reaching a maximum of 160 nmol L-1, and showed significant
correlation with phytoplankton biomass during SOAP. However, the yield of
DMS from the bacterial conversion of dissolved DMSP was variable with no spatial
trend, although a correlation with leucine incorporation indicates that DMSP
was an important carbon source for bacteria. Overall, gross DMS production
by bacteria in deck incubations of near-surface water was relatively low, at
< 6 nmol L-1 d-1, inferring that phytoplankton-mediated
conversion of DMSP was likely a significant near-surface source of DMS
(Lizotte et al., 2017).
The SSM is a potentially important interface controlling MBL and aerosol
composition, across which material exchanges between
atmosphere and ocean. Physical and biogeochemical processes within this thin
layer have the potential to alter transfer via factors, such as the
concentration of organic material and enhanced biological and photochemical
processing. Near-surface CO2 gradients have been observed (Calleja et
al., 2005), and several studies report DMS enrichment in the SSM (see
summary in Walker et al., 2016). If DMS consumption or production in the SSM
is significant, this represents a potential source of discrepancy in
comparison of measured fluxes with that calculated by the COAREG model (see
above). The biogeochemistry of the SSM and the upper 1.6 m surface water were
characterized at 10 stations during SOAP at distance from the research
vessel to determine the spatial variability in composition within, and
between, different phytoplankton blooms (Walker et al., 2016). Near-surface
DMS gradients were generally negligible, except during B1 where low
wind speed, near-surface stratification and high dinoflagellate abundance
may have enhanced DMS in the SSM relative to subsurface waters.
The observed DMS enrichment factors in the SSM during B1, ranging from 1.4
to 5.3, are some of the highest reported to date. The anomaly between
measured DMS fluxes and COAREG estimated was also greatest during B1,
inferring that DMS emissions, and associated k–wind-speed parameterizations,
may be sensitive to DMS production in the SSM under certain conditions.
However, the observations also raise questions as to how such significant
DMS enrichment is maintained in the SSM, as high DMS production would be
required to balance loss processes (Walker et al., 2016).