Introduction
Carbonyl sulfide (OCS) is the most abundant reduced sulfur compound in the
atmosphere. It enters the atmosphere either by direct emissions, e.g., from
oceans, wetlands, anoxic soils or anthropogenic emissions, or indirectly via
oxidation of the short-lived precursor gases dimethylsulfide (DMS) and carbon
disulfide (CS2) . Both precursor gases
are naturally produced in the oceans, and CS2 has an additional
anthropogenic source . With direct and
indirect marine emissions combined, the ocean is considered as the dominant source of
atmospheric OCS . The most important sink
of atmospheric OCS is uptake by terrestrial vegetation
and oxic soils, while chemical loss by
photolysis and reaction with the hydroxyl radical (OH) in the atmosphere are
minor loss processes . While tropospheric
volume mixing ratios show a distinct annual cycle , the
interannual to decadal variation is low .
Accurate accounts of sources and sinks of atmospheric OCS are crucial for two
reasons.
First, OCS is climate-relevant because it influences the radiative budget of the Earth as a greenhouse
gas and by contributing significant amounts of sulfur to the stratospheric aerosol layer
that exerts a cooling effect . The two opposite effects are currently in balance , but
future changes in atmospheric circulation, as well as the magnitude and distribution of OCS sources and sinks, could change that. Hence,
a better understanding of the tropospheric budget is needed to predict the effect of OCS in future climate scenarios .
Second, OCS has recently been suggested as a promising tool to constrain terrestrial CO2 uptake, i.e., gross primary production
(GPP), as it is taken up by plants in a similar way as CO2 . GPP, a major global CO2 flux, can only be
inferred from indirect methods, because the uptake of CO2 occurs along with a concurrent release by respiration. Unlike CO2, OCS
is irreversibly degraded within the leaf. GPP can thus be estimated based on the uptake ratio of OCS and CO2, from the leaf to regional scale
or even global scale , under the condition that other sources are negligible or well quantified. The magnitude
of terrestrial biogeochemical feedbacks on climate has been suggested to be similar to that of physical feedbacks . In order
to reduce existing uncertainties, it is thus crucial to better constrain single processes in the carbon cycle, especially GPP.
Missing source estimates derived from top-down
approaches: the listed studies used an increased vegetation sink and an a
priori direct and indirect ocean flux to estimate the magnitude of the
missing source. Assigning the missing source to oceanic emissions results in
the total ocean flux listed here. Fluxes are given in Gg S per year.
A priori
Missing
Total
Reference
ocean
source
ocean
flux
flux
235
230
465
276
600
876
289
800
1089
276
714
992
Nonetheless, current figures for tropospheric OCS sources and sinks carry
large uncertainties . While the budget has been previously
considered closed , a recent upward revision of the
vegetation sink led to a gap, i.e., a
missing source in the atmospheric budget of 230–800 Gg S per year
(Table ),
with the most recent estimates at the higher end of the range. This revision
of vegetation uptake was suggested to (i) take into account the different
deposition velocities of CO2 and OCS within the leaf and base it on GPP
instead of net primary production as well as (ii) to better
reproduce observed seasonality of OCS mixing ratios in several atmospheric
models . Based on independent top-down
approaches using MIPAS and TES satellite
observations, FTIR measurements , and NOAA ground-based
time series stations and the HIPPO aircraft campaign
, the missing source of OCS was suggested to
originate from the (tropical) ocean, most likely from the region of the
Pacific warm pool. Other potential sources such as advection of air masses
from Asia have been discussed but not tested. If the ocean
was to account for the missing source, the total top-down oceanic source
strength would then be the a priori oceanic flux plus the missing source
estimate of each inverse model simulation (Table ). This
addition would imply a 200–380 % increase in the a priori
estimated oceanic source. If oceanic direct and indirect emissions were to
account for the total missing source, an ocean source strength of
465–1089 Gg S yr-1 would be required (Table ).
OCS and its atmospheric precursors are naturally produced in the ocean. In
the surface open ocean, OCS is present in the lower picomolar range, and has
been measured on numerous cruises in the Atlantic
, including three latitudinal
transects , the Indian Ocean , the
Pacific Ocean and the Southern Ocean .
Measurements in tropical latitudes, where the missing source is assumed to be
located, have previously been performed in the Indian Ocean
and during the Atlantic transects . OCS is produced
photochemically from chromophoric dissolved organic matter (CDOM)
and by a not fully understood light-independent
production that has been suggested to be linked to radical formation
. Dissolved OCS is efficiently hydrolyzed to CO2
and H2S at a rate depending on pH and temperature . CS2
has been measured in the Pacific and Atlantic oceans in a range of
7.2–27.5 pmol L-1 and during two Atlantic transects
(summer and winter) in a range of 4–40 pmol L-1 . It
is produced photochemically and biologically
, and no significant loss process other than air–sea gas
exchange has been identified . DMS is present in the lower
nanomolar range in the surface ocean and has been extensively studied in
several campaigns, summarized in a climatology by . DMS is
biogenically produced and consumed in the surface ocean, as well as
photo-oxidized and ventilated by air–sea exchange .
Global oceanic emission estimates of OCS: direct ocean emission
estimates of OCS from bottom-up approaches. Uncertainties are given in
parentheses as in the original paper either as range or ± standard
deviation.
Reference
Emitted S as OCS
(Gg S yr-1)
Extrapolated from measurements
320 (±160)*
245*
110–210*
230 (110–210)*
160 (85–340)*
-16 (-10–30)*
41–80*
53 (±80)*
53*
Model simulations
41 (±154)
813 (573–3997)
This study
130 (±80)
* Units deviate from original
paper, converted to Gg S for comparison.
Available bottom-up estimates of the global oceanic OCS fluxes from shipboard
observations range from -16 to 320 Gg S yr-1
(Table ). However, the highest estimates were biased,
because mainly summertime and daytime observations of water concentrations
were considered. With the discovery of the seasonal oceanic sink of OCS
during wintertime and a pronounced diel cycle
, direct oceanic emissions were corrected downwards.
Only recently, OCS emissions have been estimated with the biogeochemical
ocean model NEMO-PISCES at a magnitude of
813 Gg S yr-1, sufficient to account for the missing source. This
oceanic emission inventory had been used to constrain GPP based on OCS on a
global scale . However, the oceanic OCS photoproduction in
the ocean model included a parameterization for OCS photoproduction derived
from an experiment in the North Sea , which might not be
representative for the global ocean, as indicated by photoproduction constants that were an order of magnitude
lower in the Atlantic ocean compared to the German
Bight .
Here, we present new observations in all three tropical ocean basins, two of
them measured with unprecedented precision and time resolution. Direct fluxes
were inferred from continuous OCS measurements in the tropical Pacific and
Indian oceans, covering a range of regimes with respect to CDOM content,
ultraviolet (UV) radiation and sea surface temperature (SST). These
observations are used to further constrain and validate a biogeochemical box
model which had previously been shown to reproduce OCS concentration in the
Atlantic Ocean reasonably well . The box model is now updated
from its previous global application by adding and further
developing the most recent process parameterizations to estimate the global
source strength of direct OCS emissions. The emission estimate is further
complemented by discussing the potential of indirect OCS emissions, i.e., the
emissions of short-lived precursor gases CS2 and DMS, to account for the
gap in the budget.
Methods
Study sites
Observed OCS water concentrations and calculated emissions:
observations of OCS concentrations in the surface ocean during the three cruises
(a) OASIS, (b) ASTRA-OMZ, and (c) TransPEGASO as
well as the corresponding emissions calculated based on the concentration gradient
between water and marine boundary layer (d–f). Outgassing is
indicated in red bars; oceanic uptake in blue bars. The grey line shows wind
speed measured onboard the vessels. Flux data are shown with different scales
on the y axes. Data gaps occurred during stays in port and territorial
waters or during instrument tests.
![]()
Several cruises were conducted to measure the trace gases OCS (OASIS,
TransPEGASO, ASTRA-OMZ) and CS2 (TransPEGASO, ASTRA-OMZ). Cruise tracks
are depicted in Fig. . The OASIS cruise onboard RV SONNE I to
the Indian Ocean started from Port Louis, Mauritius, to Malé, Maldives, in
July and August 2014, where mainly oligotrophic waters were encountered.
TransPEGASO was an Atlantic transect starting in Gibraltar and leading to Buenos
Aires, Argentina, and Punta Arenas, Chile. It took place in October and
November 2014 and covered a variety of biogeochemical regimes. ASTRA-OMZ
onboard RV SONNE II started in Guayaquil, Ecuador, and ended in Antofagasta,
Chile, in October 2015. Although 2015 was an El Niño year, upwelling together
with high biological production was still encountered during the cruise
.
Measurement setup for trace gases
OCS was measured during two cruises onboard the RV SONNE I (OASIS) and SONNE II (ASTRA-OMZ) with a continuous underway system similar to the one described
in , at a measurement frequency of 1 Hz. The system
consisted of a Weiss-type equilibrator, through which seawater is pumped from
approximately 5 m below the surface with a flow of 3–4 L min-1. The
air from the equilibrator headspace was Nafion-dried and continuously pumped
into an OCS analyzer (model DL-T-100, Los Gatos Research) that uses the off-axis integrated cavity output spectroscopy (OA-ICOS) technique. The instrument
used onboard is a prototype of a commercial instrument
(www.lgrinc.com/documents/OCS_Analyzer_Datasheet.pdf), developed by Los
Gatos Research (LGR) in collaboration with Forschungszentrum Jülich GmbH
. Data were averaged over 2 min, achieving a precision of
15 ppt. OCS mixing ratios in the marine boundary layer (MBL) were determined by pumping outside air
ca. 50 m from the ship's deck to the OCS analyzer (KNF Neuberger pump). A
measurement cycle consisted of 50 min water sampling and 10 min air
sampling, where the first 3 min after switching until stabilization of the
signal were discarded.
Before and after the cruise the analyzer was calibrated over a range of
concentrations using permeation devices. Both calibrations were consistent.
However, during calibration the output of the internal spectral retrieval
differed significantly from post-processing of the recorded spectra, which
matched the known concentrations (this offset is not present in the
commercial instruments). The calibration data were thus used to derive a
correction function. After correction all data stayed within 5 % of the
standards. The calibration scale of the permeation devices was 5 % below
the NOAA scale. As the OCS analyzer measured CO2 simultaneously, and
CO2 standards were available during the cruise, drift of the instrument
was tested by measuring CO2 standard gases before and after the cruise and
found to be less than 1 % of the signal. Special care was taken to avoid
contamination, and all materials used were tested for contamination
before use.
During OASIS, the mirrors inside the cavity of the OCS analyzer were not
completely clean, which led to a reduced signal. To correct the data, an
attenuation factor was determined from simultaneous CO2 measurements,
because no OCS standard was available onboard, and OASIS data were corrected
accordingly.
An independent quality check of the data was performed by comparing volume
mixing ratios of the MBL from the OCS analyzer with samples from air
canisters sampled during both cruises and measured independently
. The calibrated (and attenuation corrected for
OASIS) OA-ICOS data were on average 5 % lower than the air canister
samples, which reflects the 5 % difference between the calibration at
Forschungszentrum Jülich and the NOAA scale.
During ASTRA-OMZ, CS2 was directly measured onboard within 1 h of
collection using a purge and trap system attached to a gas chromatograph and
mass spectrometer (GC/MS; Agilent 7890A/Agilent 5975C; inert XL MSD with
triple axis detector) running in single-ion mode. The discrete surface
seawater samples (50 mL) were taken each hour to every 3 h from the
same pump system as for continuous OCS measurements. CS2 was stripped by
purging with helium (70 mL min-1) for 15 min. The gas stream was
dried using a Nafion membrane dryer (Perma Pure) and CS2 was
preconcentrated in a trap cooled with liquid nitrogen. After heating the trap
with hot water, CS2 was injected into the GC/MS. Retention time for CS2
(m/z 76, 78) was 4.9 min. The analyzed data were calibrated each day using
gravimetrically prepared liquid CS2 standards in ethylene glycol. During
purging, 500 µL gaseous deuterated DMS (d3-DMS) and isoprene
(d5-isoprene) were added to each sample as an internal standard to account
for possible sensitivity drift between calibrations.
During the TransPEGASO cruise onboard RV Hesperides, surface ocean
OCS and CS2 were measured in discrete seawater samples by purge and trap
and gas chromatography with mass spectrometry detection (GC-MSD). Samples
were collected every day at 09:00 and 15:00 local time in glass bottles
without headspace and analyzed within 1 h. Aliquots of 25 mL were withdrawn
with a glass syringe and filtered through GF/F during injection into the purge
and trap system (Stratum, Teledyne Tekmar). The water was heated to
30 ∘C and volatiles were stripped by bubbling with
40 mL min-1 of ultrapure helium for 12 min and trapped in a U-shaped
VOCARB 9 trap at room temperature. After flash thermal desorption, volatiles
were injected into an Agilent 5975T LTM GC-MSD equipped with an Agilent LTM
DB-VRX column (20 m × 0.18 mm OD × 1 µm)
maintained at 30 ∘C. Retention times for OCS (m/z 60) and CS2
(m/z 76) were 1.3 and 2.7 min, respectively. Peak quantification was
achieved with respect to gaseous (OCS in N2) and liquid (CS2 in
methanol and water) standards that were analyzed in the same way. Samples
were run in duplicates. Detection limits were 1.8 pM (OCS) and 1.4 pM
(CS2), and precision was typically around 5 %.
The systems are each calibrated against a standard, but they had not been directly
intercompared. Still, our measurements are consistent with previous
measurements using independent methods as discussed in
Sects. and .
Calculation of air–sea exchange
Fluxes F of all gases were calculated with Eq. ():
F=kw⋅ΔC,
where kw is the gas transfer velocity in water (i.e., physical
constraints on exchange) and ΔC the air–sea concentration gradient
(i.e., the chemical constraint on exchange). The air-side transfer velocity
for OCS was calculated to be 7 orders of magnitude
smaller and was therefore neglected. The concentration gradient was
determined using the temperature dependent Henry constant
and the measurements in the surface water and MBL for
OASIS and ASTRA-OMZ. During TransPEGASO, no atmospheric volume mixing ratio
was measured, and a value of 500 ppt was assumed . As air
volume mixing ratios of OCS vary over the course of a year, we performed a
sensitivity test for a scenario of 450 and 550 ppt and found mean deviations
of +7.8 and -7.8 %, respectively. The transfer velocity kw
was determined using a quadratic parameterization based on wind speed
which was directly measured onboard (10 min averages).
Furthermore, kw was corrected for OCS and CS2 by scaling it
with the Schmidt number calculated from the molar volume of the gases
. It should be noted that the choice of the parameterization
for kw has a non-negligible influence on the global emission
estimate. Linear, quadratic and cubic parameterizations of kw are
available, with differences increasing at high wind speeds on the order of a
factor of 2 . Evidence suggests that the air–sea
exchange of insoluble gases such as CO2, OCS and CS2 follows a cubic
relationship to wind speed because of bubble-mediated gas transfer
. However, this difference between soluble and
non-soluble gases is not always consistent , and too few
data are available for a reliable parameterization at high wind speeds above
12 m s-1, where the cubic and the quadratic parameterizations diverge
the most. For reasons of consistency, e.g., for the fitted photoproduction p
from previous studies, and the fact that most of the previous emission
estimates were computed using a quadratic kw parameterization, we
chose the same quadratic parameterization representing the mean range of
observations . For a sensitivity test, we computed the global
oceanic emission with a cubic relationship , which results in
an additional 40 Gg S per year as direct OCS emissions, leaving the missing
source still unexplained. However, better constraints on the transfer
velocity of insoluble gases would decrease the uncertainty in global oceanic
emissions of marine trace gases.
Box model of OCS concentration in the surface ocean
A box model to simulate surface concentration of OCS is further developed
from the latest version from termed vH2003, where
concentrations along the tracks of five Atlantic cruises have been
simulated and compared. The vH2003 model results from successful tests and
validation to observations on several cruises to the Atlantic Ocean covering
all seasons (i.e., , in January 1994;
,
in April/May 1992; , in June/July 1997; ,
in September/October 1998). By comparing photoproduction rate constants of
the five cruises to CDOM absorption, suggest a second-order
process for photoproduction with the photoproduction rate constant being
dependent on the absorption of CDOM in seawater.
In our approach, we test vH2003 along the
cruise track of two cruises, include a new way of determining the
photoproduction rate constant (see below) and apply it with global
climatological input (termed L2016). termed
K2000 applied a similar version of vH2003 globally, which
included an optimized photoproduction constant from Atlantic transect cruise
data, an optimized constant light-independent production and a linear
regression to obtain CDOM from chlorophyll a. In comparison to K2000, we
use (i) a new way of determining the photoproduction rate constant
incorporating information from three ocean basins, (ii) the most recent
parameterization of light-independent production available, and
(iii) satellite observations for sea surface CDOM instead of an empirical
relationship based on chlorophyll a.
implemented parameterizations for light-independent
production, hydrolysis and air–sea exchange similar to vH2003 in the 3-D
global ocean model NEMO-PISCES. The main differences to the approach used
here are the lack of accounting for mixing in L2016 (discussed in
Sect. , which will theoretically lead to higher simulated
concentrations in our case) and the application of a photoproduction rate
constant in our model that incorporates information from three open ocean
basins in contrast to one from a study in the North Sea .
In L2016, the light-independent production term of OCS was parameterized
depending on SST (K) and the absorption coefficient of CDOM at 350 nm
wavelength, a350 (Eq. ).
dCOCSdt=a350×10-6×exp55.8-16 200SST
An overview on symbols and abbreviations used in equations in the following is provided in the Appendix. The parameterization for hydrolysis describes alkaline and acidic degradation
of OCS by Reactions () and ():
OCS+H2O→H2S+CO2,OCS+OH-→SH-+CO2.
It was parameterized as a first-order kinetic reaction including the rate
constant kh according to Eqs. ()–():
dCOCSdt=[OCS]⋅kh,kh=exp24.3-10 450SST+exp22.8-6040SST⋅Ka[H+],-log10K=3046.7SST+3.7685+0.0035486⋅SSS,
where a[H+] is the proton activity and K the ion product of seawater
.
Fluxes were calculated with Eq. () using the same
parameterization for kw as for the emission calculation from
measurements described above.
Photoproduction was integrated over the mixed layer depth (MLD), assuming a
constant concentration of OCS and CDOM throughout the mixed layer, with the
photoproduction rate constant p (mol J-1), a350 (m-1) and UV radiation (W m-2) (Eq. ).
dCOCSdt=∫-MLD0pa350UVdz
MLD was obtained from CTD (conductivity, temperature, depth) profiles and
interpolated between these locations (Figs. S1, S2 in the Supplement). The
photochemically active radiation that reaches the ocean surface was
approximated by Eq. () :
UV=2.85×10-4⋅I⋅cos2θ,
with global radiation I (W m-2) and the zenith angle cos θ.
The attenuated UV light intensity directly below the surface
down to the respective depth of the mixed layer was calculated in 1 m steps,
taking into account attenuation by CDOM and pure seawater. As a
simplification in this global approach, the box model did not resolve the
whole wavelength spectrum, but rather used a350 and applied a
photoproduction rate constant that takes into account the integrated
spectrum. A similar approach had been tested and compared to a wavelength
spectrum resolving version by .
Box model simulations compared to observations: comparison of
simulated OCS water concentrations against measurements from the OASIS cruise
to the Indian Ocean (a) and the eastern Pacific Ocean during the
ASTRA-OMZ cruise (b). Blue indicates OCS concentrations with a
least-squares fit for the photoproduction rate constant p during daylight,
fitted individually for days with homogeneous water masses (SST, a350).
Black shows the simulation including the p depending on a350, obtained
from linear regression of individually fitted p with a350
(r=0.71). The time on the x axis is local time (GMT+5 during OASIS
2014, GMT-4 during ASTRA-OMZ 2015).
![]()
Dependence of photoproduction rate constant p on a350
including own fits for p (resulting in blue lines in Fig. 2) and fits from
a similar study . Dashed lines indicate the 95 % confidence
interval.
![]()
The rate coefficients for
hydrolysis, light-independent production and air–sea exchange are all
reasonably well constrained and parameterizations have been derived from
dedicated laboratory and field experiments (hydrolysis, air–sea exchange) or
from nighttime OCS observations in several regions assuming steady state
(dark production; ). However, the photoproduction
rate constant p is not well constrained and no generally applicable
parameterization exists. In the study of , a start was made in parameterizing
p in terms of CDOM absorption, and they found this to be dependent on the exact
model setup used with respect to wavelength integration and mixed layer
treatment. To extend the p–CDOM relationship for other ocean basins, we use
the two cruises OASIS and ASTRA-OMZ as case studies for parameter
optimization of the photoproduction rate constant p. The photoproduction
constant p in the case study simulations was fitted individually for
periods of daylight > 100 W m-2 (Fig. , blue lines)
with a Levenberg–Marquardt optimization routine in MatLab version 2015a
(8.5.0) by minimizing residuals between simulated and hourly averaged
measurements. Different starting values were tested to reduce the risk of the
fitted p being a local minimum. Together with photoproduction rate
constants obtained by a similar optimization procedure by
(Table 2 therein, termed MLB STC), a relationship of the photoproduction
constant p dependent on a350 was established (Fig. ).
The resulting linear relationship thus includes values from the Atlantic,
Pacific and Indian oceans, making it a good approximation for a globally valid
dependence. For the global box model, p was calculated in every time step
based on this relationship (r=0.71, Eq. ):
p=3591.3⋅a350+329.4
The scatter in Fig. likely reflects the inhomogeneity of the
water masses across the three oceanic basins considered, as CDOM absorbance
is a valid proxy, but carries some uncertainty in the concentration of the
actual precursor.
The model input for simulations of the cruises OASIS and ASTRA-OMZ consisted
of measurements made during the respective cruise, including SST and SSS
(MicroCAT SBE41) measured every minute, CDOM absorption coefficient
(spectrophotometrically measured ca. every 3 h with a liquid capillary cell
setup) and the ship's in situ measured meteorological data such as wind speed
and global radiation averaged over 10 min (Figs. S1, S2, Tables S1, S2).
Forcing data were linearly interpolated to the time step of integration of
2 min.
For the global box model, monthly global meteorological fields with a spatial
resolution of 2.8∘ × 2.8∘ were used (Table S3, Fig. S3). For
global a350 at the sea surface, monthly climatological means for
absorption due to gelbstoff and detritus a443 (gelbstoff representing CDOM) from the MODIS-Aqua
satellite (all available data, 2002–2014) were corrected to
350 nm with Eq. () :
a350=a443⋅exp(-0.02⋅(350-443)).
SST, wind speed, and atmospheric pressure were obtained as monthly
climatological means from the same period, i.e., 2002 to 2014, by ERA-Interim
. A diel cycle of global radiation I was obtained by fitting
the parable parameters a and b during time of the day t in
Eq. () (Fig. S4),
I=-a⋅t2+b,
to conditions of (i) x axis interceptions in the distance of the sunshine
duration and (ii) the integral being the daily incoming energy by ERA-Interim
. Monthly climatologies of mixed layer depths were used from
the MIMOC project . For details of data sources please refer
to Tables S1–S3 provided in the Supplement. The time step of the model was
set to 120 min, which had been tested to result in negligible (<3 %)
smoothing.
Assessing the indirect contribution of DMS with EMAC
Model outputs from ECHAM/MESSy Atmospheric Chemistry (EMAC) from the
simulation RC1SDbase-10a of the ESCiMo project are used to
evaluate the contribution of DMS on the production of OCS. The model results
were obtained with ECHAM5 version 5.3.02 and MESSy version 2.51, with a
T42L90MA resolution (corresponding to a quadratic Gaussian grid of approx.
2.8 by 2.8∘ in latitude and longitude and 90 vertical hybrid
pressure levels up to 0.01 hPa). The dynamics of the general circulation model
were nudged by Newtonian relaxation towards ERA-Interim reanalysis data. DMS
emissions were calculated with the AIRSEA submodel , which
takes into account concentration of DMS in the atmosphere and in the ocean,
following a two-layer conceptual model to calculate emissions
. While atmospheric concentrations are estimated online by
the model (with DMS oxidation), the oceanic concentrations are prescribed as
monthly climatologies . It was shown that such an online
calculation of emissions provides the most realistic results when compared to
measurements compared to a fixed emission rate . The online-calculated concentrations of DMS and OH were been used to estimate the
production of OCS. A production yield of 0.7 % was used for the
reaction of DMS with OH , using the reaction rate constant
suggested by the International Union of Pure and Applied Chemistry (IUPAC)
.
Results and discussion
Observations of OCS in the tropical ocean
Average, standard deviation and range of parameters observed during
the cruises OASIS (Indian Ocean, 2014), ASTRA-OMZ (Pacific Ocean, 2015) and
TransPEGASO (Atlantic Ocean, 2014).
Average (±SD)
Minimum
Maximum
OASIS (Indian Ocean)
OCS sea surface concentration
(pmol L-1)
9.1 (±3.5)
5.1
20.7
OCS flux
(g S d-1 km-2)
-0.25 (±0.5)
-1.6
1.5
SST
(∘C)
27.0 (±1.4)
22.2
32.0
Salinity
(–)
34.9 (±0.3)
34.3
35.4
Wind speed
(m s-1)
7.6 (±2.1)
0.2
14.5
aCDOM(350)
(m-1)
0.03 (±0.02)
n.d.
0.12
ASTRA-OMZ (Pacific Ocean)
OCS sea surface concentration
(pmol L-1)
28.3 (±19.7)
6.5
133.8
OCS flux
(g S d-1 km-2)
1.5 (±2.1)
-1.5
19.9
CS2 sea surface concentration
(pmol L-1)
17.8 (±8.9)
6.7
40.1
CS2 flux
(g S d-1 km-2)
4.1 (±3.2)
0.2
14.4
SST
(∘C)
20.1 (±2.9)
15.6
26.9
Salinity
(–)
35.0 (±0.43)
33.4
35.5
Wind speed
(m s-1)
7.4 (±2.0)
0.3
15.5
aCDOM(350)
(m-1)
0.15 (±0.03)
0.1
0.24
TransPEGASO (Atlantic Ocean)
OCS sea surface concentration
(pmol L-1)
23.6 (±19.3)
2.6
78.3
OCS flux
(g S d-1)
1.3 (±3.5)
-1.7
14.0
CS2 sea surface concentration
(pmol L-1)
62.5 (±42.1)
23.2
154.8
CS2 flux
(g S d-1 km-2)
13.7 (±9.8)
0.3
33.9
SST
(∘C)
22.6 (±6.3)
7.1
29.6
Salinity
(–)
34.9 (±2.6)
28.4
38.1
Wind speed
(m s-1)
7.4 (±3.1)
0.4
19.0
aCDOM(350)
(m-1)
0.13 (±0.11)
0.0023
0.45
OCS was measured in the surface ocean and MBL during three cruises in the
tropics. Measurement locations (Fig. ) include oligotrophic
open ocean regions in the Indian Ocean (OASIS, 07-08/2014), open ocean and
shelf areas in the eastern Pacific (ASTRA-OMZ, 10/2015) and a meridional
transect in the Atlantic (TransPEGASO, 10-11/2014). In the Indian and Pacific
oceans, continuous underway measurements provided the necessary temporal
resolution to observe diel cycles of OCS concentrations in surface water.
Dissolved OCS concentrations exhibited diel cycles with maxima 2 to 4 h
after local noon (Fig. ), which are a consequence of
photochemical production and removal by hydrolysis . OCS
concentrations also varied spatially. Taking a350 as a proxy for CDOM
content, we found that daily mean OCS concentrations were higher in CDOM-rich
(Table , 28.3 ± 19.7 pmol OCS L-1, a350:
0.15 ± 0.03 m-1) than in CDOM-poor waters (Table ,
OASIS: 9.1±3.5 pmol OCS L-1, a350:
0.03 ± 0.02 m-1). Samples during TransPEGASO were measured with
gas chromatography–mass spectrometry twice a day (around 08:00–10:00 and 15:00–17:00
local times). Therefore, the full diel cycles could not be reconstructed and
potential variations of OCS with CDOM absorption were overlaid by diel
variations. Nevertheless, the observed range of OCS concentrations in the
Atlantic corresponds well to the observations from the eastern Pacific and
Indian Ocean (Table ) and is consistent with measurements from
a previous Atlantic meridional transect (AMT-7) cruise
(1.3–112.0 pmol OCS L-1, mean 21.7 pmol OCS L-1).
Comparison of water properties relevant for OCS
production and consumption for the cruises OASIS (Indian Ocean,
July–August 2014) and ASTRA-OMZ (eastern Pacific,
October–November 2015) with the assumed source region in the Pacific warm
pool (15∘ N–15∘ S, 120–180∘ E). Data from
cruises are in situ measurements; the data for the Pacific warm pool were
extracted from climatological monthly means from sources for the global model
run as specified in the Supplement.
Parameter
OASIS
ASTRA-OMZ
Pacific warm pool
SST (∘C)
27.0 ± 1.0
19.6 ± 2.6
28.9 ± 0.9
SSS (g kg-1)
35.0 ± 0.3
35.1 ± 0.3
34.5 ± 0.42
Wind speed (m s-1)
8.2 ± 1.7
7.5 ± 1.8
5.3 ± 0.4
a350 (m-1)
0.039 ± 0.02
0.146 ± 0.02
0.050 ± 0.08
I (W m-2)
226.5 ± 303.0
196.4 ± 283.1
206.4 ± 286.6a
SR (J m-2)
1.9×107±1.7×106
1.6×107±4.5×106
8.9×106±1.3×106
pH (–)
8.03 ± 0.01
–b
8.07 ± 0.01
MLD (m)
43.3 ± 15.8
18.9 ± 7.5
35.9 ± 14.1
a Calculated from an annual mean diurnal cycle based
on ERA-Interim sunshine duration and flux. SR: surface radiation, daily
integral. b Assumed pH = 8.15 for box model simulation.
Air–sea fluxes calculated from surface concentrations and mixing ratios of
OCS as a function of wind speed generally follow the diel cycle of the
surface ocean concentration. While supersaturation prevailed during the day,
low nighttime concentrations usually led to oceanic uptake of atmospheric
OCS. OCS fluxes integrated over one day ranged from -0.024 to
-0.0002 g S km-2 in the open Indian Ocean and from 0.38 to
2.7 g S km-2 in the coastal Pacific. During the observed periods, the
ocean was a net sink of atmospheric OCS in the Indian Ocean, whereas it was a
net source in the eastern Pacific. Although an assessment of net flux is
difficult given the lower temporal resolution during TransPEGASO, calculated
emissions were in the same range as the ones measured in the Pacific and
Indian Ocean.
The water masses encountered during the cruises to the Indian Ocean (OASIS)
and eastern Pacific (ASTRA-OMZ), which are used to constrain the global box
model, differ considerably with respect to the properties relevant for OCS
cycling and, thus, span a large range of possible OCS variability. The
properties encountered during these two cruises encompass or exceed the ones
of the Pacific warm pool (climatological averages, Table ),
which is where the location of the missing source has been hypothesized
. Both higher SST and lower wind speeds (Table
) would decrease the OCS sea surface concentrations in the
ocean, leading to decreased emissions to the atmosphere: higher SSTs favor a
stronger degradation by hydrolysis , and lower wind speeds
decrease the transfer velocity k. Lower integrated daily radiation (SR in
Table ) in the Pacific warm pool also points to lower OCS
production. Hence, our new OCS observations presented here likely span the
range of emission variability in the tropics.
The observed concentrations and calculated emissions are approximately 1
order of magnitude lower than the annual mean surface concentrations and
emissions simulated in the 3-D global ocean model NEMO-PISCES
.
A direct global oceanic emission estimate for OCS
The OCS observations from the Indian and Pacific Ocean were used to improve a
box model for simulating OCS concentrations in the surface ocean
. With the a350-dependent
photoproduction constant included, the model reproduced the diel pattern of
OCS concentrations in the surface oceans for both cruises
(Fig. , black lines). A slight overestimation of observed
concentrations is present for the Indian Ocean cruise OASIS (observed mean
concentration: 9.1 ± 3.5 pmol L-1; simulated:
10.8 ± 3.9 pmol L-1). This overestimation was more pronounced in
the eastern Pacific (observed mean: 28.3 ± 19.7 pmol L-1;
simulated: 47.3 ± 25.4 pmol L-1) and can largely be attributed
to a lack of downward mixing inherent in the mixed layer box model due to the
assumption of the OCS concentration being constant throughout the entire
mixed layer.
Annual mean of surface ocean concentrations of OCS simulated with
the box model (a) and corresponding emissions (b).
![]()
Comparison of previous ship campaign measurements with corresponding
month and approximate geolocation from the global box model in this study
(L2016), taken either from figures or tables as provided in the original
references. Note that the box model is based on input data from
climatological means that do not fully represent the conditions encountered
during the respective cruises. Only observational data with measurements of
the full diel cycle were included for comparison. n: number of
measurements.
References
Season
Region
Mean OCS ± SD
n
L2016 mean
(pmol L-1)
(pmol L-1)
open Indian Ocean
20∘ N–37∘ S
Mar/May 1986
OCEAT II
19.9 ± 0.5a
20
11.2 ± 6.3
Jul 1987
OCEAT III
19.9 ± 1.0a
14
17.7 ± 13.1
Nov–Dec 1990
Weddell Sea
109b
126
66.6 ± 49.8
40–72∘ S,72∘ W–24∘ E
North Atlantic Ocean
Apr/May 1992
47∘ N 20∘ W
14.9 ± 6.9
118
42.8 ± 11.3
Jan 1994
48–50∘ N, 10–17∘ W
5.3 ± 1.6
120
8.9 ± 3.2
Sep 1994
48–50∘ N, 10–17∘ W
19.0 ± 8.3
235
33.4 ± 3.5
Jan 1994
northeastern Atlantic Ocean
6.7 (4–11)
120
9.6 ± 3.7
49∘ N, 12∘ W
Mar 1995
western Atlantic
8.1 ± 7.0
323
15.8
32∘ N, 64∘ W
Jun/Jul 1997
northeastern Atlantic Ocean
23.6 ± 16.0
940
30.5 ± 12.6
30–40∘ N, 8–15∘ W
Sep/Oct 1998
Atlantic transect
21.7 ± 19.1
783
22.9 ± 3.2
50∘ N–60∘ S, 1–64∘ W
Aug 1999
Sargasso Sea/BATS
8.6 ± 2.8
518
8.1
32∘ N, 64∘ W
Oct/Nov 1997
Atlantic meridional transect
14.8 ± 11.4
306
11.8 ± 12.7
53∘ N–34∘ S, 25∘ W–20∘ E
May/Jun 1998
Atlantic meridional transect
18.1 ± 16.1
440
27.8 ± 47.9
53∘ N–34∘ S, 25∘ W–20∘ E
a Converted from ng L-1 with a molar mass of
OCS of 60.07 g. b Converted from ng S L-1 with a molar
mass of S of 32.1 g.
Using the linear p-a350 parameterization for the first time in a global
model, the same box model as for the case studies is applied to estimate sea
surface concentrations and fluxes of OCS on a global scale
(Fig. ). The OCS production is consistent with the global
distribution of CDOM absorption (Fig. S5), with highest concentrations
calculated for coastal regions and higher latitudes. Despite the
photochemical hot spot in the tropics (30∘ N–30∘ S),
degradation by hydrolysis prevents any accumulation of OCS in the surface
water, as we calculated the lifetime due to hydrolysis to be only 7 h
(Fig. S5). The simulated range of water concentrations is too low to sustain
emissions in the tropics that could close the atmospheric budget of OCS
(Fig. ). With saturation ratios integrated over 1 year, the tropical ocean
(30∘ N–30∘ S) is even undersaturated with respect to OCS,
taking up 3.0 Gg S yr-1. Globally, the integration over one year
yields annual oceanic OCS emissions of 130 Gg S. Our results corroborate
the upper limit of an earlier study that used an observation-derived emission
inventory (Table ) but which includes more
process-oriented parameterizations as described in Sect. .
Clearly, our results from both observations and modeling contradict the
latest bottom-up emission estimate from the NEMO-PISCES model
, and do not support a hot spot of direct OCS emissions in
the Pacific Warm Pool or the tropical oceans in general.
Comparison to previous ship-based measurements
The global simulation of OCS surface water concentrations generally
reproduced the lower picomolar range of concentrations
(Table ), the seasonal pattern of higher concentrations
during summer compared to winter (as, for example, in ) and the
spatial pattern of higher concentrations in higher latitudes (e.g., Southern
Ocean; ). Given that monthly means of a model simulation
driven by climatological data of the input parameters are compared to cruise
measurements, the absolute mean deviation of 6.9 pmol L-1 and the mean
deviation of 3.7 pmol L-1 indicate an overall good reproduction of
observations (differences between observation and model output were weighted
to the number of observations in Table ). It should be noted that,
on average, the model overestimates OCS concentrations as indicated by the
positive mean error, suggesting our emission estimate to be an upper limit to
direct oceanic OCS emissions in most regions. The largest deviations from
observations are found in the Southern Ocean (see , in
Table ), where the model underestimated observations by 40 %.
While there are several explanations for this, i.e., a possible violation of
the underlying assumption of a constant OCS production in regions with deep
mixed layers such as the Southern Ocean, or the missing satellite data for
CDOM during polar nights, it is a clear indication of the need for more
observations from high latitudes. However, this underestimation does not
interfere with our conclusion drawn for the tropical oceans, where the
location of the missing source is derived from top-down approaches.
Uncertainties
Simulated concentrations and fluxes carry uncertainties from input parameters
and process parameterizations. One major uncertainty associated with the
mixed layer box model approach arises from the fact that it does not
adequately account for downward mixing and vertical concentration gradients
within the mixed layer. Under most circumstances, and especially in the
tropical open ocean, where hydrolysis greatly exceeds surface outgassing and
low a350 makes photoproduction extend further down in the water column,
the model tends to overestimate the real OCS concentrations, as was shown for
our two cruises above. Therefore, we deem the fluxes from our global
simulation to represent an upper limit of the true fluxes. Only at high
latitudes would we expect more complex uncertainties, because hydrolysis at
low temperatures is slow and only photoproduction and loss by outgassing are
directly competing at the very surface.
Other uncertainties are associated with the calculation of the
photoproduction rate. The wavelength of 443 nm combines the absorption of
detritus and CDOM, which could have an impact especially in river plumes,
where terrestrial material is transported into the ocean. As it is the CDOM
that is important for photochemistry, assuming the 443 nm is purely CDOM
would lead to an overestimation of photoproduction and therefore is a
conservative estimate. It should also be noted that a single spectral slope
from 443 to 350 nm in the global simulation is a simplification.
Furthermore, using a wavelength integrated photoproduction rate constant
instead of a wavelength-resolved approach, which would take global variations
in the CDOM and light spectra into account, is an additional simplification.
It has been shown that this does not lead to large differences regionally
but could, potentially, lead to variations globally. Our
p–CDOM relationship is a first step for constraining this variability
globally in one parameterization, as it incorporates photoproduction rate
constants optimized to observations and thus accounting for differences in
the light and CDOM spectra. More data from different regions can help to
further constrain this relationship in future studies. Despite these
simplifications, the simulated concentrations agree very well with previous
observations (n>4000, Table ). To test the sensitivity of our
box model to the photoproduction rate constant, we performed a sensitivity
test with a photoproduction increased by a factor of 5 in the tropical region
(30∘ N–30∘ S; note that this factor is considerably larger
than the uncertainty in the p–CDOM relationship). This leads to an annual
mean concentration of 35.1 pmol L-1 in the tropics
(30∘ N–30∘ S), resulting in tropical direct emissions of
160 Gg S as OCS per year. The efficient hydrolysis in warm tropical waters
prevents OCS concentrations from accumulating despite the high
photoproduction and still results in emissions too low to account for the
missing source.
With a mean error of 3.7 pmol L-1 in the OCS surface
water concentrations added to (subtracted from) the modeled concentration
and subsequent calculation of fluxes using annual climatologies for wind,
pressure and SST (same data sources as global simulation forcing data), we
calculate an uncertainty of 60 %, which translates into a total uncertainty
in the integrated global flux of 80 Gg S yr-1.
Indirect OCS emissions by DMS and CS<inline-formula><mml:math display="inline"><mml:msub><mml:mi/><mml:mn mathvariant="normal">2</mml:mn></mml:msub></mml:math></inline-formula>
Measured concentration of CS2 in surface waters during
(a) ASTRA-OMZ in the eastern Pacific Ocean and (b) TransPEGASO
in the Atlantic Ocean.
![]()
A significant contribution to the OCS budget in the atmosphere results from
oceanic emissions of DMS and CS2 that are partially converted to OCS on
timescales of hours to days . A yield of
0.7 % for OCS is used for the reaction of DMS with OH ,
which results in a global oceanic source of DMS from OCS of 80 (65–110) Gg S yr-1 based on the procedure described in
Sect. . The uncertainty range of 65–110 Gg S yr-1
originated from the uncertainty in oceanic emissions, not the conversion
factor. This conversion factor is much more uncertain, as the formation of
OCS from DMS involves a complex multi-step reaction mechanism that is not
fully understood. It has been shown in laboratory experiments that the
presence of NOx reduces the OCS yield considerably , which
would make our indirect emission estimate an upper limit. However, the yield
was measured under laboratory conditions and may be different and more
variable under natural conditions.
DMS emissions do not show a pronounced hot spot in the Pacific warm pool
region, but as DMS transports much more sulfur across the air–sea interface
than OCS, even low changes in the OCS yield could affect the atmospheric
budget of OCS. As the spatial oceanic emission pattern of DMS does not
reflect the spatial pattern of the assumed missing source, a locally specific
tropospheric change in the conversion yield would be one potential way of
bringing the patterns in agreement. While it is possible that the OCS yield
could vary under certain conditions (e.g., it cannot be excluded that the low
OH concentrations in the broader Pacific warm pool area as suggested by
, influence the yield), the (local) increase in the conversion
factor would need to be on the order of a factor of 10–100.
For CS2, the atmospheric reaction pathway producing OCS is better
understood with a well-constrained molar conversion ratio of 0.81
. However, the global distribution of oceanic CS2
concentration, and hence its emissions to the atmosphere, is poorly known. In
our study, surface CS2 concentrations (Fig. S6) were on average
17.8 ± 8.9 pmol L-1 during ASTRA-OMZ, and
62.5 ± 42.1 pmol L-1 during TransPEGASO (Table ).
The latter values are higher than previously reported concentrations from the
AMT-7 cruise in the central Atlantic
(10.9 ± 15.2 pmol L-1). We extrapolate a weighted mean of the
CS2 emissions from TransPEGASO (n=42,
13.7 ± 9.8 g S d-1 km-2), ASTRA-OMZ (n=122,
4.1 ± 3.2 g S d-1 km-2) and AMT-7
(n=744, 1.6 ± 1.8 g S d-1 km-2) in order to estimate
CS2-derived OCS emissions from the global ocean. According to our
extrapolation, 135 (7–260) Gg S yr-1 enters the atmosphere as oceanic
CS2 emissions converted to OCS. The uncertainty range of
7–260 Gg S yr-1 results from extrapolating the highest and the
lowest emissions encountered during the cruises to the global ocean. This
number is at the highest end of the range for OCS emissions from
globally simulated CS2 oceanic concentrations ,
as measured CS2 concentrations from the cruises ASTRA-OMZ and TransPEGASO
are higher than the simulated surface concentrations in for
the respective month. However, the spatial pattern of higher concentrations
and emissions in the tropical region in our measurements agrees well with the
spatial pattern simulated in . Nonetheless, even the
extrapolation of the highest measurement would not close the budget for the
three largest missing source estimates (Table ).
For
oceanic emission estimates used to constrain GPP, quantifying the seasonal
cycle of the single contributors is essential. For example, high emissions
during oceanic spring and fall blooms could mask OCS uptake by the
terrestrial vegetation, and therefore neglecting them could lead to an
underestimation of global GPP, with implications for the atmospheric and
terrestrial carbon budget.
Conclusions and outlook
Considering the observational evidence
and the modeled global emission estimate of
130 ± 80 Gg S yr-1, direct OCS emissions from the oceans are
too low to account for the missing atmospheric source. Together with indirect
emissions, the oceanic source strength of OCS would add up to
345 Gg S yr-1, compared to the 465–1089 Gg S yr-1 required to
balance the suggested increase in vegetation uptake. Direct and even
additional indirect oceanic emissions of OCS are thus unlikely to balance the
budget after the upward revision of the vegetation sink. Largest
uncertainties are associated with the indirect emission estimates, especially
in the conversion of DMS to OCS and the global source strength of CS2.
As our study suggests, the search for an additional source of OCS to the
atmosphere should include other sources than oceanic emissions alone. There
are indications of other parts of the OCS budget being underestimated, such
as domestic coal combustion . Emissions of biomass burning and
direct and indirect anthropogenic emissions have been considered in previous
estimates (e.g., 315.5 Gg S yr-1 in ,
224 Gg S yr-1 in , and 219 Gg S yr-1 in
), but a recent anthropogenic emission estimate by
increases this number to 598 Gg S yr-1, which would
already bring sources and sinks closer to agreement. They attribute the
largest direct OCS emissions to biomass and biofuel burning, as well as pulp
and paper manufacturing, and the largest CS2 emissions to the rayon
industry. Hence, a hot spot of anthropogenic emissions in the Asian continent
might be a potential candidate, together with atmospheric transport, to
produce atmospheric mixing ratios as observed by satellite.
A redistribution of the magnitude and seasonality of known sources and sinks
could also bring top-down and bottom-up estimates closer together. For
example, the general view of oxic soils as a sink for OCS has recently been
challenged. Field and incubation studies
show that some oxic soils may shift from OCS uptake to
emission depending on the temperature and water content. Furthermore, it has
been speculated previously that vegetation uptake might not be solely
responsible for the decrease in OCS mixing ratios in fall because of the
temporal lag between CO2 and OCS minimum . The observed
seasonality in mixing ratios is a superposition of the seasonality of all
individual sources and sinks. These seasonalities are currently neglected or
associated with a considerable uncertainty. An improved understanding of the
seasonality of the individual sources and sinks could help to better
constrain the gap in the atmospheric budget. First steps to resolve OCS
seasonality in sources and sinks are currently being undertaken, e.g., in the case
of anthropogenic emissions .
All in all, better constraints on the seasonality and magnitude of the
atmospheric OCS sources and sinks are critical for a better assessment of the
role of this compound in climate and its application to quantify GPP on a
global scale. This study confirms oceanic emission as the largest known
single source of atmospheric OCS but shows that its magnitude is unlikely to
balance the gap in the atmospheric OCS budget.