Photolysis and the dimethylsulfide (DMS) summer paradox in the Sargasso... Dierdre A. Toole

Photolysis and the dimethylsulfide (DMS) summer paradox in the Sargasso... Dierdre A. Toole
Limnol. Oceanogr., 48(3), 2003, 1088–1100
q 2003, by the American Society of Limnology and Oceanography, Inc.
Photolysis and the dimethylsulfide (DMS) summer paradox in the Sargasso Sea
Dierdre A. Toole
Institute for Computational Earth System Science; Interdepartmental Graduate Program in Marine Science,
University of California–Santa Barbara, Santa Barbara, California 93106
David J. Kieber
State University of New York, College of Environmental Science and Forestry, Chemistry Department,
Syracuse, New York 13210
Ronald P. Kiene
Department of Marine Sciences, University of South Alabama, Mobile, Alabama 36688
David A. Siegel and Norman B. Nelson
Institute for Computational Earth System Science, University of California–Santa Barbara,
Santa Barbara, California 93106
Apparent quantum yields and rates of dimethylsulfide (DMS) photolysis were determined from Sargasso Sea
seawater with the goal of assessing the extent to which photoreactions affect the unusually elevated upper ocean
concentrations of DMS during the summer, the so-called DMS summer paradox. Apparent quantum yields determined with monochromatic radiation decrease exponentially with increasing wavelength and indicate that DMS
photolysis is driven by ultraviolet (UV) radiation. The relative spectral partitioning differs between samples collected
from the surface mixed layer (15 m) and from the chlorophyll a maximum (80 m), presumably because of differences
in chromophoric dissolved organic matter (CDOM) quality (e.g., apparent degree of bleaching). Quantum yields
are also temperature dependent, and an approximate doubling of photolysis rates occurs for a 208C increase in
temperature. The significance of DMS photolysis to upper ocean sulfur budgets is explored using a multiyear (1992–
1994) DMS time series, concurrent irradiance determinations and temperature profiles, and estimates of CDOM
absorption. Depth-integrated, mixed-layer DMS photolysis rates peak in the summer (15–25 mmol m22 d21) and
decline to ,1 mmol m22 d21 in the winter. These rates correspond to specific turnover rates of ;0.29 d21 in the
summer and ,0.02 d21 in the winter. Seasonal changes in solar radiation, temperature, and DMS concentrations
drive the 30-fold differences in photolysis rates, overshadowing differences caused by photosensitizer (CDOM)
quantity or quality (21–35%). These results demonstrate that although photolysis is not the primary driver of the
summer paradox, it makes an important contribution to the time–depth pattern of DMS concentrations observed in
the Sargasso Sea.
The biogeochemical cycling of sulfur between the upper
water column of the ocean and the atmospheric marine
boundary layer has received a great deal of attention over
the last several decades because of its implication in a cloud
albedo feedback loop (e.g., Shaw 1983; Bates et al. 1987;
Charlson et al. 1987). The biogenic production of dimethylsulfide (DMS) in the marine environment is an important
source of atmospheric sulfur. Although DMS concentrations
exhibit considerable spatial and temporal variability, DMS
This work was supported by NASA under an Earth System Science Fellowship and the SIMBIOS Program. We are extremely
grateful to John Dacey for the use of his DMS and DMSP time
series datasets. The authors acknowledge George Westby and Todd
Medovich (SUNY-ESF) for their assistance with laboratory photolysis studies; Rod Johnson, Paul Lethaby, Rachel Parsons, Chrissy
Van Hilst, Karen Paterson, and the many BATS technicians for their
assistance at sea; and Andrew Hall, Toby Westberry and two anonymous reviewers for their assistance with the development of this
is ubiquitous in the ocean, representing .90% of the oceanic
sulfur flux and .50% of the global biogenic flux to the
atmosphere (Andreae 1986). DMS is derived from the enzymatic cleavage of dimethylsulfoniopropionate (DMSP),
which is synthesized in phytoplankton at species-dependent
concentrations (Keller et al. 1989). Bulk water column DMS
concentrations are a result of the balance between bacteria-,
phytoplankton-, and zooplankton-mediated production
mechanisms and loss mechanisms that primarily consist of
sea–air flux, bacterial consumption, and photolysis (e.g.,
Dacey and Wakeham 1986; Kiene and Bates 1990; Kieber
et al. 1996; Simó and Pedrós-Alió 1999b). Because of the
complex coupled physical, biological, and chemical processes involved in the marine sulfur cycle, an assessment of
the patterns of DMS variability requires a complete understanding of the substances and mechanisms associated with
each pathway.
Photolysis is an important removal process for upper
ocean DMS (e.g., Brimblecombe and Shooter 1986; Kieber
et al. 1996; Brugger et al. 1998; Hatton 2002). Photolysis
occurs through a secondary photosensitized pathway because
DMS photolysis and the summer paradox
DMS does not absorb light at wavelengths .260 nm (Brimblecombe and Shooter 1986). The exact composition and
activity of the chromophoric photosensitizers remains unclear, however, and total absorption by chromophoric dissolved organic matter (CDOM) can be used as a proxy (Kieber et al. unpubl.). Hence, photolysis can be modeled as a
function of an apparent quantum yield and the photon flux
absorbed by CDOM. From the northern Adriatic Sea, Brugger et al. (1998) showed that DMS photolysis was proportional to the irradiance intensity and dissolved organic carbon (DOC) concentrations. DOC and CDOM concentrations
are not related in the open ocean, however (Siegel and Michaels 1996; Nelson et al. 1998; Siegel et al. 2002). It is
critical to make this distinction because the colored and noncolored components of the dissolved organic matter pool
have distinct, independent, annual cycles (Nelson and Siegel
2002). Accurate photolysis parameterizations will greatly enhance current DMS models and, as a result, our understanding of DMS cycling. At present, photolysis rates are poorly
constrained, assumed constant in space and time, or simply
omitted (e.g., Van den Berg et al. 1996; Simó and PedróAlió 1999b; Jodwalis et al. 2000).
The multiyear Dacey et al. (1998) DMS time series collected as part of the U.S. Joint Global Ocean Flux Study
(JGOFS) Bermuda Atlantic Time-series Study (BATS) demonstrates the so-called summer paradox (Simó and PedrósAlió 1999a). The essence of the paradox is that DMS concentrations reach their maximum throughout the water
column in July and August, more than 2 months later than
the maximum for its precursor, DMSP (Fig. 1A,B). The timing of the DMS maximum is additionally puzzling because
it coincides with the seasonal minimum in phytoplankton
pigment biomass and primary production. Summer mean
mixed-layer chlorophyll concentrations are ;0.05 mg m23,
indicating extremely oligotrophic conditions (Fig. 1C). Mean
mixed-layer DMS concentrations also have an inverse relationship with CDOM absorption (Fig. 1E). This relationship
led Siegel and Michaels (1996) to hypothesize that the summertime buildup of DMS results from reduced photolysis
rates. Although incident ultraviolet (UV, 280–400 nm) and
photosynthetically available radiation (PAR, 400–700 nm)
fluxes are larger in the summer (Fig. 1D), the concurrent,
extremely low CDOM concentrations (Fig. 1E) suggest lower absorbed quanta, and thereby photolysis rates, in the upper
meter (Siegel and Michaels 1996). Knowledge of the functional dependences of the apparent quantum yield is required
to properly assess how the photon flux absorbed by CDOM
translates to the photochemical destruction of DMS in the
photic zone.
Here, we determine wavelength and temperature-dependent apparent quantum yields for DMS photolysis in the Sargasso Sea. Broadband solar and monochromatic incubation
experiments are conducted to determine the spectral regions
primarily responsible for DMS photolysis in this open ocean
region. The seasonal importance of DMS photolysis is explored using the Dacey at al. (1998) DMS dataset and the
extensive time series datasets collected in conjunction with
the BATS and Hydrostation S sampling programs, allowing
the controlling factors to be elucidated. A primary goal in
Fig. 1. Time series of (A) mean mixed-layer DMS concentrations (nmol L21), (B) mean mixed-layer total (dissolved 1 particulate) DMSP concentrations (nmol L21), (C) mean mixed-layer Chl
a concentrations (mg m23), (D) daily integrated surface photosynthetically available radiation (PAR) (mol quanta m22 d21), and (E)
modeled chromophoric dissolved organic matter absorption coefficient at 325 nm (m21). aCDOM(443) was modeled using the globally
optimized semianalytical UCSB IOP inversion model (Maritorena
et al. 2002) and extended into the UV using the functional relationship described in Eq. 1. DMS and DMSP 1992–1993 data are from
Dacey et al. (1998). Data are shown from January 1992 through
November 1994.
this analysis is to quantify the role that photolysis plays in
driving the observed summer paradox in the Sargasso Sea.
Materials and methods
Shipboard sampling—The northern Sargasso Sea has a
well-documented seasonal cycle characterized as seasonally
oligotrophic with a variable-strength spring bloom. The onset of warming in the spring leads to a well-stratified water
column with a shallow mixed layer in the summer that progressively deepens via convective overturn to mixed-layer
depths in excess of 100 m in the winter (Steinberg et al.
2001). Water samples were collected from the BATS site
(water depth ;4,680 m), nominally located 85 km southeast
of the island of Bermuda. Samples were collected from the
surface mixed layer (15 m) and the chlorophyll a (Chl a)
maximum (80 m) at 1030 h local time during a U.S. BATS
Toole et al.
in the pretreated 4-liter Qorpak bottles wrapped in aluminum
foil to eliminate light exposure. In addition, all samples were
drawn wearing powder-free polyethylene gloves to further
avoid contamination. All refrigerated samples were transported back to the laboratory in Syracuse, New York, within
4 d of sampling and were reaerated prior to analysis with
ultrahigh-purity air passed through high-capacity molecular
sieve, drierite, and activated carbon cartridges.
Fig. 2. CDOM absorption coefficient spectra for the 15- and
80-m samples (m21). Data collected from the diode array were used
for wavelengths 280–450 nm, with the exception of 350–370 nm,
which were smoothed using the functional relationship described in
Eq. 1.
core cruise (BATS 154) on 19 July 2001 aboard the R/V
Weatherbird II. At this time of the year, because of seasonal
stratification, surface waters are characterized by severely
photobleached CDOM (Nelson et al. 1998). Reflecting this,
the surface, 15-, and 25-m samples all had CDOM absorption coefficients ,0.225 m21 at a wavelength of 300 nm. We
chose to study the 80-m sample because its absorption coefficient was significantly higher (.0.300 m21, Fig. 2). This
depth region (60–85 m) is characterized by a seasonal maximum in UV light absorption because of local production of
CDOM (Nelson et al. 1998).
Seawater samples were gravity filtered directly from Niskin bottles secured to the conductivity, temperature, and
depth (CTD) rosette system. The samples were filtered using
a Whatman POLYCAP 75 AS 0.2-mm filter capsule (nylon
membrane with a glass microfiber prefilter enclosed in a
polypropylene housing) attached to the Niskin bottles with
silicon tubing. Prior to sample collection, filter capsules were
rinsed with Milli-Q water and acetonitrile (Burdick and Jackson, distilled in glass grade) until the fluorescence at 400 nm
was zero as determined by flow-injection analysis (Miller
2000). Unless otherwise noted, all glassware used in this
study, including the Qorpak bottles and their Teflon-lined
caps, was cleaned by three rinses of Milli-Q water followed
by three rinses each of 10% hydrofluoric acid, 10% hydrochloric acid, and 10% methanol (all chemicals were reagent
grade or better). As a final step, the glassware was copiously
rinsed with Milli-Q water to remove any chemical residue.
The cleaned bottles were then baked in a muffle furnace at
3508C for at least 6 h and subsequently stored with a small
amount of 10% hydrochloric acid prior to sampling. All
glassware was rinsed copiously with the 0.2-mm-filtered seawater sample at least five times before shipboard sample
collection. The filtered seawater samples were stored at 48C
Chemicals and standards—The DMS used for photochemical experiments was purchased from Aldrich, and
DMS standards were made from DMSP hydrochloride purchased from Research Plus. Reagent-grade sodium hydroxide purchased from Fisher was used to convert DMSP hydrochloride into DMS. For details regarding calibrations and
DMS additions to seawater for the photochemical experiments, the reader is referred to Kieber et al. (1996).
Calibrated DMS standards (;2 mmol L21 in Milli-Q water) were added to 0.2-mm-filtered seawater to yield DMS
concentrations ranging from 15 to 20 nmol L21. DMS photolysis in seawater follows pseudo first-order kinetics at concentrations up to ;50–100 nmol L21, above which it may
approach zero-order kinetics (Kieber et al. 1996). Care was
taken to stay well within the first-order concentration range.
Laboratory analysis—DMS concentrations were determined using a modified purge and trap method following
Kiene and Service (1991). Briefly, 1 ml of sample was
sparged for 2 min for analysis on a Shimadzu GC-14A gas
chromatograph (GC) equipped with a flame photometric detector and a Chromosil 330 column (Supelco). The GC was
operated at a column temperature of 608C and a detector
temperature of 2258C. All samples were analyzed in duplicate with analytical precision on the order of 62% at the 15
nmol L21 range.
CDOM absorption values were determined using a Hewlett Packard 8453 UV-visible (UV-VIS) photodiode array
spectrophotometer equipped with a 5-cm quartz microliter
flow cell, with Milli-Q water as the reference. To assess
changes associated with sample transport, CDOM absorption
values determined in Syracuse were compared to absorption
spectra obtained with freshly collected samples in Bermuda
using a Perkin Elmer Lambda 18 spectrophotometer
equipped with matched 10-cm quartz cuvettes (Nelson et al.
1998). No statistically significant changes in CDOM absorption coefficients were observed relative to the initial absorption (,5% differences were noted for the entire absorption
spectrum). Absorption curves were corrected for offsets (assumed zero between 675 and 700 nm) and fit using a standard CDOM exponential functional relationship outlined in
Green and Blough (1994).
aCDOM (l) 5 aCDOM (l0 )e 2S(l2l0 )
l is the wavelength, l0 is the reference wavelength (320 nm),
aCDOM(l) is absorption by CDOM at wavelength l, aCDOM(l0)
is CDOM absorption at 320 nm, and S is the exponential
slope parameter derived from the slope of the natural log–
transformed data versus wavelength from 290 to 350 nm.
DMS photolysis and the summer paradox
Monochromatic incubations—Wavelength-dependent and
temperature-dependent apparent quantum yields for DMS
photolysis were determined according to the procedure outlined in Kieber et al. (1996). Briefly, apparent quantum
yields were determined employing monochromatic irradiations with a 1,000-W xenon arc lamp and a Spectral Energy
GM 252 high-intensity grating monochromator. All irradiations were carried out in a 1-cm quartz cell sealed with a
Teflon-lined plastic cap with no headspace and containing a
Teflon-covered stir bar. For each irradiation, the quartz cuvette was rinsed three times and was then allowed to overflow equivalent to two or three cell volumes to rinse out the
previous sample. In the sample holder, the cell temperature
was controlled by a Fisher 9008 recirculating ethylene glycol
water bath. With the exception of the temperature study, all
incubations were carried out at 218C, and samples were allowed to equilibrate in the cell holder for 5 min prior to
irradiation. The length of the irradiation varied from 20 to
960 min depending on wavelength and sample absorption.
The bandwidth was set to 9.9 nm for wavelengths ,360 nm
and was increased to 19.5 nm for wavelengths $360 nm. In
addition, for wavelengths .320 nm, a long-pass filter (l ,
;300 nm were blocked) was placed between the monochromator and the sample to eliminate second-order diffraction.
The irradiation wavelength and bandwidth were periodically
verified employing a calibrated, Optronics OL 754 scanning
At the conclusion of an irradiation, the Teflon stopper was
pierced twice with PEEK tubing. One tube was used to draw
the sample from the bottom of the cuvette, and the other
tube served as ventilation to prevent the creation of a vacuum. The sample was immediately injected into the sparge
tube of the GC. Separate samples for initial DMS concentration were drawn from the same quartz tube used to fill
the cell just prior to and following the GC analysis of the
incubated sample.
For each water sample, from 10 to 12 wavelengths were
examined, and each wavelength was irradiated at least in
duplicate. The amount of DMS photolyzed during an irradiation ranged from 3 to 25% of the initial DMS concentration. Parallel dark samples were examined to account for any
DMS loss associated with the experimental setup. Leakage
through the Teflon stopper and other dark processes accounted for a DMS loss of ;2% over 16.5 h. Most incubation
times ranged from 20 to 180 min, so dark loss was negligible, but all quantum yield determinations were corrected
for this loss.
Once photolysis rates were determined, quantum yields
were then calculated. Absolute quantum yields could not be
determined, however, because as stated previously, the photosensitizers for DMS photolysis are not known. Therefore,
the apparent quantum yield was determined as
d(DMS) l
FPO (l) 5
Pl (1 2 10 2A l )
where FPO(l) is the wavelength-dependent apparent quantum
yield for DMS photolysis, d(DMS)l /dt is the wavelengthspecific photolysis rate for DMS (mol DMS L21 s21), V is
the volume of seawater irradiated (liters), Pl is the radiant
flux (mol photons s21), and Al is the absorbance, with (1 2
102Al) representing the fraction of incident radiation absorbed by CDOM. The photon flux in the 1-cm quartz cell
was determined using nitrite chemical actinometry (Jankowski et al. 1999; Jankowski et al. 2000). The apparent
quantum yield for photolysis represents a minimum estimate
of the true quantum yield, since it is likely that only a fraction of CDOM photosensitizes the photolysis of DMS in
Polychromatic rooftop incubations—To verify the wavelength regions primarily responsible for DMS photolysis,
polychromatic rooftop incubations were performed using
ambient sunlight and broadband cutoff filters. The 15- and
80-m seawater samples were exposed to solar radiation on
7–8 (15 m) and 11–12 August 2001 (80 m) on the roof of
the laboratory in Syracuse, New York. Specifically designed,
Teflon-sealed quartz tubes (Kieber et al. 1997) containing
;20 nmol L21 DMS in 0.2-mm-filtered seawater were placed
in a shallow, circulating, freshwater bath at 208C. The water
bath was placed on a flat black background to minimize
scattered or reflected light. Duplicate quartz tubes were subjected to different spectral irradiances using long-pass filters
of Mylar D polyester film and UF3-plexiglas with cutoffs at
313 and 400 nm (see Miller 2000, for transmission spectra),
approximating a UV-B (280–320 nm) and a total UV filter.
respectively. DMS photolysis in these spectral treatments
was compared to samples exposed to the entire solar spectrum in quartz tubes and to dark controls wrapped in aluminum foil. Initial DMS concentrations were determined in
all samples prior to rooftop deployment. Quartz tubes were
rotated in the water bath hourly to minimize differences due
to sample location. CDOM absorption spectra were determined in all quartz tubes initially and at the conclusion of
each incubation.
Quantum yield determinations—For both the 15- and
80-m seawater samples, wavelength-dependent apparent
quantum yields decreased exponentially with increasing
wavelength in the UV (Fig. 3A,B). The quantum yield spectra were accurately described by simple exponential functions determined from nonlinear least squares regression
(290–380 nm, r 2 5 0.96 for 15 m; 290–410 nm, r 2 5 0.98
for 80 m).
FPO (l, 15 m) 5 (DMS)31.9e 20.0499l
FPO (l, 80 m) 5 (DMS)0.10e 20.0321l
The 15-m sample, characterized by lower CDOM absorption, had significantly higher apparent quantum yields (at the
95% CI) in the UV-B compared with the 80-m sample. This
indicates that the mixed-layer sample (15 m) is more effective on a per-photon-absorbed basis for shorter UV wavelengths (,320 nm), whereas the deeper sample (80 m) is
more effective at longer UV wavelengths (.360 nm, Fig.
3C). Although this spectral shape is fairly characteristic of
photochemical processes in seawater (e.g., Neale and Kieber
Toole et al.
ple at 320 nm was a linear function of light dose up to 90
min, after which, rates were nonlinear, presumably because
of photobleaching of the DOM photosensitizer(s) (data not
shown). In addition, to ascertain whether bleaching of
CDOM was important during monochromatic irradiations,
seawater samples were irradiated with no added DMS, at the
same wavelengths and time intervals used for apparent quantum yield incubations, to monitor changes in the CDOM
absorption coefficients. For both the 15- and 80-m samples
and for all wavelengths, no statistically significant changes
in CDOM absorption were observed relative to dark controls
(,5% differences were noted for the entire absorption spectrum). Because CDOM absorption coefficients did not
change during monochromatic irradiations, it was assumed
that concentrations of the components of CDOM related to
DMS photolysis also did not significantly change. For both
samples and at all wavelengths, quantum yields were determined in the linear region where reciprocity was observed
(i.e., apparent quantum yields were constant).
Fig. 3. Wavelength-resolved apparent quantum yields of DMS
photolysis for (A) the 15-m sample and (B) the 80-m sample (mol
DMS photolyzed per mol photon absorbed by CDOM per nmol
DMS). In panels A and B, the circles and crosses, respectively, are
replicates and the solid line and dashed lines, respectively, are the
best fit exponential curves. (C) Mean spectral DMS quantum yields
for the 15- and 80-m samples (symbols remain the same). Determinations were made at 218C.
2000), the observed slopes are high compared to apparent
quantum yield slopes for other photochemical reactions in
the open ocean. These slopes range from 0.0250 to 0.0271
nm21 for the production of carbon monoxide (Nelson unpubl.
data), 0.0120 to 0.0272 nm21 for the production of hydrogen
peroxide (Miller 2000; Yocis et al. 2000), 0.0398 nm21 for
the production of carbon disulfide (Xie et al. 1998), and
0.0288 nm21 for the production of carbonyl sulfide (Weiss
et al. 1995). The slopes of the quantum yield functions
(0.0499 and 0.0321 nm21 for the 15- and 80-m samples,
respectively) were also much steeper than the corresponding
slopes for the CDOM absorption coefficients (290–350 nm,
0.0208 and 0.0181 nm21 for the 15- and 80-m samples, respectively). This indicates that on a per-photon basis, shorter
UV-absorbing chromophores are more effective in driving
DMS photolysis than longer wavelength-absorbing chromophores.
Reciprocity studies were conducted to insure that the incubations for the quantum yield determinations were sampled during the period where photolysis is a linear function
of photon exposure (Neale and Kieber 2000). For these experiments, separate seawater samples were irradiated for specific time intervals, generally ranging between 0 and 90 min.
For example, the photolysis of DMS in the 80-m water sam-
Temperature dependence—In addition to its wavelength
dependence, the rate of DMS photolysis was temperature
dependent. Temperature dependence studies were conducted
on both water samples at 320 nm and at temperatures ranging from 4.9 to 34.58C. DMS photolysis rates increased exponentially with increasing temperature following simple
Arrhenius behavior (Fig. 4A,B). Simple exponential relationships derived from nonlinear least squares regression described the temperature dependence (r 2 5 0.96 for 15 m; r 2
5 0.98 for 80 m).
5 0.00165e 0.0319T
5 0.00185e 0.0351T
T is temperature (8C), and the pre-exponential constant corresponds to a DMS concentration of 1 nmol L21. Over the
temperature range studied, the quantum yield increased by a
factor of 2.72 and 2.83 for the 15- and 80-m samples, respectively, resulting in an approximate doubling of quantum
yields for an increase of 208C. Activation energies for the
photolysis of DMS, determined from the Arrhenius plots,
were 22.7 6 1.2 and 24.8 6 1.9 kJ mol 21 for the 15- and
80-m samples, respectively (Fig. 4A,B insets). For modeling
purposes, although it was not explicitly verified here, these
temperature relationships were assumed constant across all
Polychromatic rooftop experiments—Photolysis rates for
both the 15- and 80-m samples were greatest in the untreated
quartz tubes as compared to the Mylar and UF3-plexiglas
treatments. On the basis of differences between light treatments, maximum photolysis rates were observed in the
UV-A (320–400 nm). In the 15-m sample, characterized by
lower CDOM absorption coefficients, the UV-B was responsible for 38.8% of the photolysis, with UV-A wavelengths
accounting for 61.2% (Fig. 5A). The 80-m sample, with the
higher CDOM absorption, was ;50% less sensitive to UV-B
radiation, with UV-B wavelengths accounting for only
DMS photolysis and the summer paradox
Fig. 4. Temperature dependence of DMS photolysis rates for
(A) the 15-m sample and (B) the 80-m sample. Rates are scaled to
an initial DMS concentration of 1 nmol L21. The inset Arrhenius
plots are the natural log of calculated quantum yields derived from
the temperature dependence experiments versus 1,000 divided by
temperature in Kelvin.
20.4% of the photolysis and UV-A responsible for the remaining 79.6% (Fig. 5B). In addition, over the duration of
the 2-d incubation, approximately three times as much DMS
was photolyzed in the 80-m sample versus the 15-m sample.
Unfortunately, concurrent irradiance measurements were not
available during these incubations. To eliminate this as a
source of the observed differences, new aliquots of the 15and 80-m samples were exposed to solar radiation on the
same days, focusing solely on the UV wavelength region
(14–15 August 2001). These incubations yielded quantitatively similar results to those observed in the initial rooftop
irradiation experiments and confirmed the higher rates in the
80-m sample, consistent with the higher CDOM absorption
coefficients and higher apparent quantum yields in the UV-A
(Fig. 3). Rates were approximately 2.5-fold higher in the
80-m sample, and the UV-B and UV-A wavelength regions
were responsible for 35.5 and 64.5% and 19.2 and 80.8% of
the DMS photolysis for the 15- and 80-m samples, respectively. Although some degree of nonlinearity is expected because the production and loss of DMS photolysis precursors
does not necessarily parallel that of bulk CDOM, the observed wavelength partitioning was likely a result of differences in photosensitizer concentrations, as suggested by the
Fig. 5. Percentage of initial DMS photolyzed during the polychromatic broadband rooftop experiments using natural sunlight for
(A) the 15-m sample and (B) the 80-m sample. All treatments were
in duplicate, and the inlaid pie charts are approximate spectral partitioning of UV-B (280–320 nm) and UV-A (320–400 nm) contributions to the total photolysis based on differences between the
various light treatments. UV-B accounted for 38.8 and 20.4% and
UV-A for 61.2 and 79.6% of the total photolysis for the 15- and
80-m samples, respectively.
higher CDOM absorption coefficients in the 80-m water
sample (Fig. 2), and photosensitizer quality, as reflected in
initial slope factors (0.0208 nm21 for 15 m vs. 0.0181 nm21
for 80 m) (Green and Blough 1994; Nelson et al. 1998).
For all rooftop incubations, the differences between the
dark replicates and the replicates incubated under UF3-plexiglas were statistically insignificant (,5%) indicating that
DMS photolysis does not occur at wavelengths greater than
;400 nm in Sargasso Sea seawater (Fig. 5). This finding
differs from previous results from the Pacific which, employing identical techniques to determine action spectra as
those used in the present study, showed significant photolysis rates in the visible portion of the solar spectrum (Kieber
et al. 1996). Sample storage artifacts associated with Pacific
seawater cannot be ruled out, however, since the 0.2-mmfiltered Pacific samples used in laboratory studies were
stored refrigerated in Teflon-lined polyethylene bottles for
more than a month. Although this is unlikely, as shipboard
photolysis experiments using freshly collected samples
showed no attenuation of photolysis rates in polycarbonate
enclosures with a 365-nm cutoff, it might have contributed
to the observed differences between the Sargasso Sea and
Pacific action spectra.
Toole et al.
Photolysis rates—As mentioned previously, the depth-dependent DMS photolysis rate, PO(z), is defined as the product of the photolysis quantum yield FPO and the absorbed
quanta AQ(z, l, t),
PO(z) 5
FPO (l)AQ(z, l, t) dt dz dl
FPO (l)aCDOM (z, l)E0 (z, l, t)
dt dz dl
where aCDOM(z, l) is the spectral CDOM absorption coefficient, E0(z, l, t) is the depth-dependent spectral scalar irradiance integrated over 1 d, h is Planck’s constant, c is the
speed of light in a vacuum, and l/hc converts the photon
flux at a given wavelength to energy. Following Eq. 7, depthdependent photolysis rates were determined by combining
the monochromator-determined quantum yields, measured
CDOM absorption coefficients, and a measure of the spectral
scalar irradiance. In this case, a monthly UV climatology
determined from atmospheric radiative transfer calculations
(including column-integrated ozone abundance and climatological cloud properties) was applied (see Lubin et al.
1998). The downward irradiance was converted to scalar irradiance assuming a conversion factor of 1.2. On the basis
of underwater radiative transfer simulations for the upper 5
m, a solar zenith angle of 308, clear skies, and the wavelength range of maximum photolysis activity (300–380 nm),
the conversion factor varied from 1.15–1.17 in the summer
to 1.16–1.24 in the winter. For completely diffuse conditions,
the conversion factor ranged from 1.21–1.24 in the summer
to 1.22–1.29 in the winter. These simulations were carried
out for extreme cases in the Sargasso Sea; thus, these ranges
can be considered outer bounds for the planar to scalar irradiance conversion factor at this site. Although this factor
does vary as a function of wavelength, depth, sky state, and
the inherent optical properties of the water column, 1.2 is a
reasonable average value (HYDROLIGHT, Mobley 1994;
Nelson unpubl. data).
Using these calculations, DMS photolysis rates were determined for the upper cubic meter of the water column and
as a function of depth in the water column (see Discussion).
Modeled surface DMS photolysis rates were extremely low
at 290 and 300 nm because of the low solar flux at these
wavelengths and increased to a maximum at 320 nm, followed by a decrease with increasing wavelength (Fig. 6).
Although the 80-m sample is characterized by lower quantum yields at shorter wavelengths (,360 nm), when the
same scalar irradiance was applied, the higher CDOM absorption dominated, resulting in a larger photolysis rate for
all wavelengths. Integration of the area under the curves
shown in Fig. 6 indicated that the photolysis rate for the
80-m samples was 175% higher than that determined for the
15-m sample, approximating results obtained in the rooftop
studies, which suggested a 2.5-fold difference between samples. Additionally, integration showed that UV-B and UV-A
accounted for 32.6 and 67.4% of the photolysis for the 15-m
sample and 22.2 and 77.8% of the photolysis for the 80-m
sample. This spectral partitioning of the active photolysis
Fig. 6. Wavelength-resolved surface photolysis rates for the
15- and the 80-m seawater samples. Photolysis rates were determined using Eq. 7, mean measured quantum yields, sample-specific aCDOM(l), and a regionally specific climatological UV incident
flux (Lubin et al. 1998). Quantum yield spectra were scaled to an
initial DMS concentration of 1 nmol L21 and 218C. Pie charts are
the approximate spectral partitioning of UV-B (280–320 nm) and
UV-A (320–400 nm) contributions to the total photolysis based
on integration of the photolysis rates. UV-B accounted for 32.6
and 22.2% and UV-A for 67.4 and 77.8% of the total photolysis
for the 15- and 80-m samples, respectively.
region matched that observed in the rooftop experiments under natural solar radiation (38.8 and 61.2% for the UV-B
and UV-A, respectively, in the 15-m samples and 20.4 and
79.6% for the 80-m sample), verifying our monochromator
Seasonal cycle modeling—The Dacey et al. (1998) DMS
time series (1992–1994) provides a clear example of the
summer paradox (Fig. 1A–C). To assess the role of photolysis in this phenomenon, photolysis rates were examined in
space and time in conjunction with the various forcing factors. Following Eq. 7, depth-integrated photolysis rates were
determined by combining (1) the laboratory-measured apparent quantum yields of photolysis, (2) spectral CDOM absorption coefficients as a function of depth, (3) spectral incident irradiance and diffuse attenuation coefficients as a
function of depth, and (4) the monthly resolved time series
of DMS concentrations. All photolysis calculations were integrated to the base of the seasonal mixed layer derived from
same-day CTD casts and over one day length. The calculations implicitly assumed that the concentration of DMS
and the CDOM absorption coefficients remained constant
over 1 d.
DMS photolysis and the summer paradox
Table 1. Estimated seasonal DMS photolysis rates and specific turnover rates. Photolysis rates are integrated to the base of the surface
mixed layer. Summer is defined as 1 May–30 Sep.
Photolysis rate
(mmol DMS m22 d21)
Turnover rate
E0(l, z)
BATS reconstructed
BATS reconstructed
Lubin et al. climatology
Lubin et al. climatology
UCSB inversion model
1995–2001 BATS climatology
UCSB inversion model
1995–2001 BATS climatology
CDOM absorption coefficients, incident UV flux, and underwater UV profiles were not taken concurrently with the
DMS time series and were therefore modeled. As a sensitivity analysis to assess how the calculation procedures affected
the final photolysis rates, and thus the overall robustness of
the results, several parallel calculations were carried out.
These calculations included two independent CDOM absorption coefficient parameterizations and two incident irradiance parameterizations. Results of this comparison are
shown in Table 1. Furthermore, DMS apparent quantum
yields (Eqs. 3, 4) were scaled to temperature (Eqs. 5, 6) and
ambient DMS concentrations. Separate model calculations
were performed using the apparent quantum yield spectra
from the 15- and 80-m samples to mimic the range of
CDOM qualities expected in the Sargasso Sea.
CDOM modeling—CDOM absorption coefficients were
modeled using a 7-yr mean of measurements at this site and
derived from a semianalytical ocean color inversion model.
Beginning in 1995, as part of the Bermuda Bio-Optics project (BBOP) program, CDOM measurements were collected
approximately 16 to 20 times a year in the upper 140 m
(Nelson et al. 1998). Although there is considerable interannual variability, a depth-resolved seasonal composite of
all available data from 1995–2001 was created, yielding the
well-documented temporal and spatial patterns (Nelson et al.
1998; Nelson and Siegel 2002). CDOM absorption is generally lowest in the summer in the mixed layer when stratification is strongest and bleaching occurs. It is highest at
depth in the summer because of high local production rates
and is fairly homogeneous with depth in the winter because
of physical mixing processes, low biological activity, and
low UV fluxes (Siegel and Michaels 1996; Nelson et al.
CDOM absorption was also modeled using the globally
optimized semianalytical University of California–Santa
Barbara (UCSB) inherent optical property (IOP) ocean color
inversion model (Maritorena et al. 2002). Using remote sensing reflectances at wavelengths 412, 443, 490, 510, and 555
nm, this model estimates surface CDOM and detrital absorption at a wavelength of 443 nm, particulate backscattering at a wavelength of 443 nm, and Chl a concentrations.
Underwater profiles of downwelling irradiance and upwelling radiance were determined from near-noon optical casts
sampled by the Marine Environmental Radiometer package
(Biospherical Instruments; Siegel et al. 2001). The optical
profiles were extrapolated to and across the air–sea interface,
providing remote sensing reflectances. CDOM absorption
was extended spectrally using Eq. 1 and a spectral slope of
0.0206 nm21, which was found to produce the most consistent results when tuned globally (Maritorena et al. 2002). In
this case, CDOM absorption was considered constant within
the mixed layer.
In both CDOM parameterizations, the absorption coefficients were considered constant over the time course of 1 d.
Although the concentration of chromophoric DMS photolysis precursors might change during this time period as the
reaction progresses, without knowledge of the exact reaction
photosensitizers or the reaction mechanism, it is impossible
to model this change. Nelson et al. (1998) demonstrated that
bulk CDOM in the Sargasso Sea has a lifetime of approximately 90 d with respect to photobleaching processes and
14 d with respect to bacterial production processes, suggesting that this was a reasonable assumption.
Irradiance modeling—The incident dose (i.e., time-integrated irradiance) was modeled via (1) a literature UV climatology and (2) irradiance reconstructed from measured
visible fluxes. The first method consisted of using a spectral
incident irradiance 18 3 18 monthly UV climatology (Lubin
et al. 1998). The second downwelling UV flux estimate was
constructed from shipboard measurements of nearly concurrent (62 d) measured daily integrated visible irradiances.
Relying on climatological results is not prudent because this
site experiences considerable deviation in daily incident irradiance because of cloud variability (Siegel et al. 2001; Fig.
1D). During core BATS cruises from August 2000 to November 2001, a deck-mounted cosine-corrected radiometer
sampled throughout the day, allowing for the development
of empirical relationships between daily integrated UV and
visible irradiances. Ratios between integrated UV irradiance
and integrated visible irradiance at 412 nm were surprisingly
constant throughout the 15-month period, at 0.305 6 0.017
(SD) for 324 nm, 0.451 6 0.016 for 340 nm, and 0.542 6
0.015 for 380 nm. Applying these ratios to daily integrated
Ed(412) measurements from the original sampling period, a
time series of integrated irradiance at 324, 340, and 380 nm
was reconstructed. Extrapolation to other UV wavelengths
was based on inter-UV wavelength ratios derived from the
Lubin et al. (1998) climatology. These ratios were constant
throughout the year and matched up with similar ratios from
the BBOP 2000–2001 dataset quite well. For example, the
Lubin et al. (1998) climatology has a mean Ed(324)/Ed(340)
ratio of 0.703 6 0.012, whereas the mean BBOP Ed(324)/
Ed(340) ratio was 0.677 6 0.022. This reconstruction allowed the modeled incident light fields to more accurately
Toole et al.
reflect the solar conditions on the day the time series seawater samples were collected.
Depth-dependent irradiance was estimated from diffuse attenuation coefficients, which can be related to the absorption
and scattering coefficients (Kirk 1981).
Kd(l) 5 Ïa(l) 2 1 0.231a(l)b(l)
Kd(l) is the downwelling diffuse attenuation coefficient, a(l)
is the spectral total absorption coefficient, and b(l) is the
spectral total scattering coefficient. Total scattering coefficients were derived as a function of chlorophyll concentrations (Fig. 1C) using a basic bio-optical model (Gordon and
Morel 1983). Absorption by pure water was extended into
the UV assuming that the spectral slope for wavelengths
380–410 nm remained constant to 290 nm (using data from
Pope and Fry 1997). During core BATS cruises from August
1999 to December 2001, profiles of UV irradiance and radiance were sampled, allowing for the development of an
empirical absorption parameterization. It was found that approximating total absorption as the sum of absorption by
pure water and absorption by CDOM determined from the
Maritorena et al. (2002) retrievals, extended spectrally with
a lower S factor (0.0194 m21) (Eq. 1), combined with scattering following Eq. 8 yielded the most accurate Kd(l) spectrally (modeled vs. measured Kd(l), r 2 5 0.94 for l # 412
nm, slope 5 1.001, intercept 5 20.003, n 5 33 profiles).
Seasonal photolysis rates and turnover times—Apparent
quantum yield spectra, along with modeled CDOM absorption, modeled underwater irradiance fields, and the time series of DMS concentrations and temperature were used to
calculate depth-integrated photolysis rates following Eq. 7.
Figure 7A shows an example of mixed-layer integrated photolysis rates based on the CDOM absorption derived from
the ocean color inversion estimates and the incident UV reconstructed from BATS data (Table 1, case 1). Peak summertime photolysis rates were greater than 16 mmol DMS
m22 d21, whereas midwinter photolysis rates were ,1 mmol
DMS m22 d21. Corresponding specific turnover rates in the
summer ranged from 0.05 to 0.22 d21 and as low as 0.01
d21 in the winter (Fig. 7B). Use of the climatological aCDOM(z,
l) lead to relatively higher photolysis rates and a proportional increase in specific turnover rates because of a variety
of factors (Table 1, case 2, 4 vs. case 1, 3). First, the ocean
color inversion parameterization does not take into account
seasonal variations in the CDOM absorption exponential
slope parameter S and assumes that one factor holds
throughout the entire UV wavelength region. Log-transformed CDOM absorption spectra deviate from linearity in
the shorter UV wavelengths, suggesting a larger S factor for
shorter wavelengths and thus a relative underestimation by
the ocean color inversion model. In addition, because of the
high degree of interannual variability associated with CDOM
absorption measurements, a 7-yr mean will not necessarily
be representative of the actual CDOM absorption during
1992–1994. Use of the Lubin et al. (1998) monthly climatological UV dataset tended to slightly overestimate the
downwelling irradiance relative to the BATS reconstructed
Fig. 7. (A) Mixed-layer, depth-integrated DMS photolysis rates
for the 15- and 80-m sample, (B) specific turnover rates in the
surface mixed layer calculated as photolysis rate divided by mixedlayer integrated DMS concentration, (C) total absorbed quanta
(290–400 nm) that is the product of aCDOM(l) and E0(l, z), and (D)
contours of depth-resolved absorbed quanta at 325 nm overlaid with
mixed-layer depth (thick solid line). These calculations were carried
out using the globally optimized semianalytical UCSB IOP inversion model (Maritorena et al. 2002) and incident UV reconstructed
from the BATS visible dataset (Table 1, case 1).
estimates, producing increased photolysis rates and larger
specific turnover times (Table 1, case 3, 4 vs. case 1, 2).
Qualitatively, the results are robust; the different calculations all indicate the same seasonal pattern. For all calculations, the difference between the summer maximum and
winter minimum in DMS photolysis rates was approximately
30-fold, whereas differences from the use of the two apparent quantum yields was only 20 to 35%. Most likely, the
mixed-layer photolysis rates resemble those derived from the
15-m quantum yield in the summer, when the water column
is extremely stratified, and those calculated from the 80-m
quantum yield in the winter, when deep convective mixing
results in deeper sources of higher absorbing CDOM.
The pattern of DMS photolysis rates observed in Fig. 7
is primarily due to seasonal changes in DMS concentrations,
seawater temperature, and the incident irradiance. Because
DMS photolysis is a pseudo–first-order reaction, wavelength-dependent quantum yields are directly proportional to
ambient DMS concentrations, which vary by a factor of .12
DMS photolysis and the summer paradox
seasonally (Fig. 1A). The quantum yields also scale to the
in situ temperature following Eqs. 5 and 6. Sea surface temperatures are cold (19–228C) and homogeneous with depth
in the winter and warm to a maximum of 27–308C during
the spring and the summer. If all other factors remained constant, the increase in temperature from the winter to the summer would result in an ;40% increase in summer DMS
photolysis rates.
Mixed-layer integrated absorbed quanta were also slightly
higher in the summer, reflecting the balance between increased incident irradiance, lower CDOM absorption, and
shallower mixed-layer depths (Fig. 7C). Essentially, the absorbed quanta are the product of the fraction of light absorbed by CDOM and the incident irradiance in the mixed
layer and will thus be affected by seasonal variations in the
intensity and spectral distribution of actinic solar radiation
in the water column. In the Sargasso Sea, CDOM absorption
coefficients are roughly two to three times lower in the summer (Fig. 1E) as the CDOM bleaches through UV exposure,
reducing its photosensitizing ability and permitting deeper
penetration of UV radiation. At the same time however, the
incident UV irradiance concurrently increases two to three
times depending on meteorological conditions (Fig. 1D).
While the mixed-layer depth mirrors the seasonal temperature progression, ranging from .150 m in the winter to
depths as shallow as 15–20 m in the summer, its effect on
integrated absorbed quanta, and thus photolysis rates, is minimal (Fig. 7D). In the summertime, although UV wavelengths are not necessarily 100% absorbed in the surface
mixed layer, because of the exponential attenuation of light,
an appreciable percentage are (Fig. 7D). During the winter,
when mixed layers are very deep (.150 m), UV wavelengths are essentially 100% absorbed. Thus, there is very
little seasonal difference in the degree to which UV radiation
is absorbed within the surface mixed layer. Seasonal differences in the integrated absorbed quanta are primarily dominated by seasonal variations in the availability of UV irradiance damped by lower CDOM absorption. The result is
slightly elevated, integrated mixed-layer absorbed quanta in
the summertime, driving higher DMS photolysis rates.
Seasonal changes in CDOM absorption coefficients will
affect the depth horizon over which UV photons are attenuated, however. Radiometer casts (SPMR, Satlantic) taken
concurrently with the initial seawater sampling indicated that
the 1% light level for 324 nm, approximately the wavelength
of maximum photolysis, was 42 m, and the 1% light level
for 412 nm was 93 m (Fig. 8). Summertime in the Sargasso
Sea is characterized by some of the optically clearest seawater worldwide, so these 1% light levels can be considered
upper bounds for the depth over which photolysis occurs. In
general, even in the clearest open ocean waters, UV radiation
is exponentially attenuated in the upper 20 or 30 m, and
consequently, the majority of absorbed quanta and water column DMS photolysis occurs at these shallow depths. As
shorter wavelengths of light are attenuated preferentially, the
spectral composition of in situ UV light will shift to longer
wavelengths. Therefore, as depth increases, fewer total photons are present and there is a simultaneous shift to longer
wavelengths characterized by lower quantum yields and lower CDOM absorption. The net result is that fewer photons
Fig. 8. Attenuation of downwelling irradiance, Ed(l) with
depth, from the cast just prior to water sample collection on 19 July
2001 (wavelengths are noted next to each irradiance profile). One
percent light levels are noted with square markers. Radiometric
measurements were collected using a SeaWiFS profiling multichannel radiometer (SPMR).
are absorbed and less DMS photolysis occurs per photon
Implications for the summer paradox—Calculations presented here demonstrated that DMS photolysis peaks in the
summer, concurrent with larger incident UV irradiance, temperatures, and in situ concentrations of DMS. They also indicate that photolysis is not the main factor controlling the
DMS summer paradox. If photolysis controlled mixed-layer
DMS concentrations, one would expect DMS concentrations
to be lowest during the summer and this is exactly opposite
of what is observed in this region. Photolysis contributes to
the dampening of DMS concentrations in the upper water
column (,20 m) but cannot explain the large midsummer
buildup deeper in the water column. Clearly, other mechanisms must lead to the elevated DMS concentrations in the
summer. Seasonal changes in wind-driven sea–air losses of
DMS cannot explain this trend because atmospheric ventilation is at a maximum in the summer because of increased
DMS concentration gradients between the ocean and the
boundary layer and increased oceanic temperatures (see discussion below). Chemical oxidation of DMS is also negligible because DMS has residence times on the order of years
with respect to that process (Shooter and Brimblecombe
1989). Ultimately, with abiological losses eliminated as the
primary drivers of the observed pattern of DMS concentrations, there are two, potentially interacting, light-driven processes driving the summer paradox: either a reduction in
bacterial consumption of DMS (e.g., Herndl et al. 1993) or
an increase in DMS production by bacteria or phytoplankton
(e.g., Sunda et al. 2002) or both. To gain an insight into the
observed pattern and forcing mechanisms, we can draw from
Toole et al.
many of the previous studies that have focused on specific
elements of the sulfur cycle and their relationship with light.
One possibility for the summer paradox is a reduction in
summertime bacterial DMS consumption rates. Ultraviolet
radiation is capable of transforming refractory DOM into
more labile forms stimulating bacterial activity (Kieber
2000). In addition, CDOM is produced locally as a by-product of microbial metabolism (Nelson et al. 1998; Nelson and
Siegel 2002). Seasonal changes in CDOM concentration and
composition can lead to changes in bacterial activity that
presumably affect DMS bacterial consumption (Kiene and
Bates 1990). Experimentally, Slezak et al. (2001) demonstrated that, under surface irradiance conditions, microbial
removal of DMS was inhibited by 59.6 6 14.0% (6SD)
versus dark controls. They found that various spectral regions had differential inhibitory effects on bacterial consumption of DMS, supporting the importance of the spectral
quality of light as modulated by CDOM absorption. In addition, although bacterial abundance is rather invariable, certain subpopulations are highly seasonal, making it difficult
to generalize literature consumption rates. Supporting this,
Slezak et al. (2001) noted high spatial and temporal variability in bacterial sensitivity to solar irradiation. Depending
on the quantity and quality of CDOM, the spectral shape of
the downwelling irradiance, and the speciation and sensitivity of the bacterial population, a buildup of DMS could occur
as bacterial consumption is inhibited by the seasonal increase in the clarity of the water column from the winter to
the summer.
Under conditions of inhibition from increased UV radiation, bacterial sulfur demands could be less, suggesting
changes in how DMSP is processed. Bacterial populations
assimilate DMSP as a source of reduced sulfur for protein
synthesis (e.g., Kiene et al. 1999). Kiene et al. (2000) found
that the microbial conversion rate is dependent on DMSP
concentration and bacterial demand. They observed that at
higher DMSP concentrations, a lower fraction of DMSP was
assimilated, with a larger portion being converted to DMS
as bacterial sulfur demands were satisfied. Bacterial sulfur
demands depend on the biomass, growth rate, and composition of the bacterial community in the water column. Their
community composition and activity can be affected to considerable depths, however, depending on the light regime,
further complicating direct interpretation of past results (e.g.,
Herndl et al. 1993; Simó and Pedrós-Alió 1999a).
Another plausible explanation for the DMS summer paradox is that increased UV radiation in the summertime results in an increase in the phytoplankton conversion rate of
DMSP to DMS. Changes in either the spectral intensity or
composition of incident radiation will affect the balance between the biological production of DMSP by phytoplankton
and the subsequent conversion to DMS. A summertime increase in DMS in conjunction with increased water clarity
and irradiance is consistent with the antioxidant function of
DMSP and DMS demonstrated by Sunda et al. (2002). As
oxidative stressors such as UV radiation increase, intracellular phytoplankton production of DMSP and its resulting
enzymatic cleavage to DMS is expected to increase. Sunda
et al. (2002) found that the greatest accumulation of DMS
and DMSP per cell volume was observed at intermediate
levels of UV stress. At higher levels of UV stress (full UV-B
exposure), the overall metabolic production of DMSP and
DMS is expected to be greater, but the oxidative removal
should also be faster, resulting in lower net cellular DMSP
concentrations relative to levels observed under intermediate
UV stress. Although it might result from a variety of concurrent factors, the midwater column maximum (20–40 m)
in DMS evident in the time series is consistent with the
antioxidant hypothesis.
Increased DMSP and DMS production in the summer
might also be explained through the seasonal evolution in
the phytoplankton community and the food web structure
(e.g., Simó and Pedrós-Alió 1999a; Simó et al. 2002). The
seasonal evolution in phytoplankton community composition
in the deep chlorophyll maximum layer at BATS has a characteristic pattern of spring bloom conditions followed by a
diverse community with increased contributions from prymnesiophytes in the spring and early summer, prochlorophytes
and cyanobacteria in late spring to early fall, and pelagophytes increasing in late summer (e.g., Steinberg et al. 2001).
Dinoflagellates and prasinophytes make a small contribution
annually. At this depth horizon, the phytoplankton community structure does shift to species associated with higher
DMSP production, but this production is not evident in the
time series of DMSP stocks (Fig. 1B). In addition, in highly
oligotrophic regions, the food web is dominated by recycling
organisms, providing more opportunity for processing of
Ultimately, the importance of DMS photolysis as a loss
mechanism is gauged by comparing the photolysis-specific
turnover rates to the specific turnover rates of sea–air flux
and bacterial consumption. There have been startlingly few
studies in open ocean regions that have attempted to concurrently sample all relevant parameters to make this assessment. At our study site, using standard flux parameterizations (Liss and Merlivat 1986) and concurrently measured
wind speed, sea–air flux-specific turnover rates in the surface
mixed layer range from ;0–0.18 d21 (mean 5 0.03 d21) in
the winter to ;0–0.31 d21 (mean 5 0.04 d21) in the summer.
Although there are days where sea–air losses can be quite
high because of isolated wind events, the low mean specific
turnover rates indicate that, seasonally, photolysis is a larger
DMS sink.
Bacterial consumption rates typically show a high degree
of variability in open ocean sites (e.g., Kiene and Bates
1990; Kieber et al. 1996; Ledyard and Dacey 1996; Wolfe
et al. 1999). Ledyard and Dacey (1996) found net bacterial
consumption rates in the surface mixed layer that ranged
from 0.02 to 0.15 nmol L21 h21 for this site during the winter
and were negligible in the summer. These bacterial consumption rates can be applied to the Dacey et al. (1998)
depth-integrated DMS concentrations to yield estimated bacterial specific turnover rates. The consumption rates were
assumed constant throughout the mixed layer, yielding turnover rates ranging from 0.09–1.4 d21 (assuming 0.02 nmol
L21 h21) to 0.69 to .5 d21 (assuming 0.15 nmol L21 h21) in
the winter. These are likely to be maximal rates, however,
because Ledyard and Dacey’s (1996) incubations were conducted in amber Qorpak bottles, which do not allow for photoinhibition. In addition, the fact that they observed no ap-
DMS photolysis and the summer paradox
preciable net bacterial consumption in the summer, along
with the previously mentioned bacterial consumption inhibition, suggests significantly longer biological turnover times
during time periods of high UV flux.
In absolute terms, because we did not concurrently measure bacterial consumption, we cannot assess the relative importance of photolysis to total DMS loss during the study
period. A recent study indicates that under typical surface
irradiation conditions, DMS photolysis is always more rapid
than bacterial consumption (Slezak et al. 2001). In a shortterm Lagrangian study in the subpolar North Atlantic, Simó
et al. (1999b) estimated all loss pathways in the mixed layer
and demonstrated that for the first 5 d, characterized by clear
skies, photolysis was the dominant DMS sink, whereas for
the remaining 9 d, characterized by lower solar radiation
because of clouds, bacterial consumption dominated. Although the importance of these two loss terms will vary as
a function of irradiance, water clarity, temperature, and bacterial activity and sulfur demands, there are clearly time periods and depth regions where each process will dominate.
It is clear from this study that DMS photolysis significantly limits the buildup of upper ocean DMS but does not
drive the summer paradox in the Sargasso Sea. Moreover,
light has a complex and variable role in the biogeochemical
cycling of sulfur on adaptations and inhibition of individual
species and communities. DMS and DMSP contribute a
large portion of the organic sulfur fluxes and a significant
portion of the total carbon flux through primary and secondary producers (e.g., Kiene et al. 2000; Burkill et al. 2002;
Simó et al. 2002). Knowledge of these food web interactions
is critical for developing a predictive understanding of DMS
cycling and its exchange with the atmosphere. The potential
radiative and climatic ramifications resulting from a decoupling of production and loss of DMS further necessitate understanding the spatial and temporal variations in the relevant mechanisms that control DMS concentrations in the
upper water column. Without a more complete understanding of the complex and dynamic food web interactions and
a way to independently measure production and removal
terms, it is difficult to determine the origins of the summer
ANDREAE, M. O. 1986. The ocean as a source of atmospheric sulfur
compounds, p. 331–362. In P. Buat-Menard [ed.], The role of
air–sea exchange in geochemical cycling. Academic.
BATES, T. S., R. J. CHARLSON, AND R. H. GAMMON. 1987. Evidence
for the climatic role of marine biogenic sulphur. Nature 329:
BRIMBLECOMBE, P., AND D. SHOOTER. 1986. Photo-oxidation of dimethylsulphide in aqueous solution. Mar. Chem. 19: 343–353.
1998. Photolysis of dimethylsulfide in the northern Adriatic
Sea: Dependence on substrate concentration, irradiance and
DOC concentration. Mar. Chem. 59: 321–331.
S. B. GROOM, G. A. TARRAN, AND M. V. ZUBKOV. 2002. Dimethyl sulphide biogeochemistry within a coccolithophore
bloom (DISCO): An overview. Deep-Sea Res. 49: 2863–2885.
WARREN. 1987. Oceanic phytoplankton, atmospheric sulfur,
cloud albedo and climate. Nature 326: 655–661.
DACEY, J. W. H., AND S. G. WAKEHAM. 1986. Oceanic dimethylsulfide: Production during zooplankton grazing on phytoplankton. Science 233: 1314–1316.
Temporal variability of dimethylsulfide and dimethylsulfoniopropionate in the Sargasso Sea. Deep-Sea Res. 45: 2085–2104.
GORDON, H. R., AND A. MOREL. 1983. Remote assessment of ocean
color for interpretation of satellite visible imagery, a review;
lecture notes on coastal and estuarine studies, v. 4. SpringerVerlag.
GREEN, S. A., AND N. V. BLOUGH. 1994. Optical absorption and
fluorescence properties of chromophoric dissolved organic
matter in natural water. Limnol. Oceanogr. 39: 1903–1916.
HATTON, A. D. 2002. Influence of photochemistry on the marine
biogeochemical cycle of dimethylsulphide in the northern
North Sea. Deep-Sea Res. 49: 3039–3052.
HERNDL, G. J., G. MULLER-NIKLAS, AND J. FRICK. 1993. Major role
of ultraviolet-B in controlling bacterioplankton growth in the
surface layer of the ocean. Nature 361: 717–719.
JANKOWSKI, J. J., D. J. KIEBER, AND K. MOPPER. 1999. Nitrate and
nitrite ultraviolet actinometers. Photochem. Photobiol. 70:
, AND P. J. NEALE. 2000. Development and
intercalibration of ultraviolet solar actinometers. Photochem.
Photobiol. 71: 431–440.
JODWALIS, C. M., R. L. BENNER, AND D. L. ESLINGER. 2000. Modeling of dimethyl sulfide ocean mixing, biological production,
and sea-to-air flux for high latitudes. J. Geophys. Res. 105:
Dimethyl sulfide production in marine phytoplankton, p. 167–
182. In E. Saltzman and W. J. Cooper [eds.], Biogenic Sulfur
in the Environment. American Chemical Society Symposium
Series No. 393. Am. Chem. Soc., Washington, D.C.
KIEBER, D. J. 2000. Photochemical production of biological substrates, p. 130–148. In S. J. de Mora, S. Demers, and M. Vernet
[eds.], The effects of UV radiation in the marine environment.
, J. JIAO, R. P. KIENE, AND T. S. BATES. 1996. Impact of
dimethylsulfide photochemistry on methyl sulfur cycling in the
equatorial Pacific Ocean. J. Geophys. Res. 101: 3715–3722.
, B. H. YOCIS, AND K. MOPPER. 1997. Free-floating drifter
for photochemical studies in the water column. Limnol. Oceanogr. 42: 1829–1833.
KIENE, R. P., AND T. S. BATES. 1990. Biological removal of dimethylsulfide from sea water. Nature 345: 702–705.
, AND S. K. SERVICE. 1991. Decomposition of dissolved
DMSO and DMS in estuarine waters: Dependence on temperature and substrate concentration. Mar. Ecol. Prog. Ser. 76: 1–
, L. J. LINN, J. GONZALEZ, M. A. MORAN, AND J. A. BRUTON. 1999. Dimethylsulfoniopropionate and methanethiol are
important precursors of methionine and protein-sulfur in marine bacterioplankton. Appl. Environ. Microbiol. 65: 4549–
, AND J. A. BRUTON. 2000. New and important roles
for DMSP in marine microbial communities. J. Sea Res. 43:
KIRK, J. T. O. 1981. Estimation of the scattering coefficient of natural waters using underwater irradiance measurements. Aust.
J. Mar. Freshw. Res. 32: 333–339.
LEDYARD, K. M., AND J. W. H. DACEY. 1996. Microbial cycling of
DMSP and DMS in coastal and oligotrophic seawater. Limnol.
Oceanogr. 41: 33–40.
Toole et al.
LISS, P. S., AND L. MERLIVAT. 1986. Air–sea gas exchange rates:
Introduction and synthesis, p. 113–127. In P. Buat-Menard
[ed.], The role of air–sea exchange in geochemical cycling.
LUBIN, D., E. H. JENSEN, AND H. P. GIES. 1998. Global surface
ultraviolet radiation climatology from TOMS and ERBE data.
J. Geophys. Res. 103: 26,061–26,091.
MARITORENA, S., D. A. SIEGEL, AND A. R. PETERSON. 2002. Optimization of a semi-analytical ocean color model for globalscale applications. Appl. Opt. 41: 2705–2714.
MILLER, G. W. 2000. Wavelength and temperature dependent quantum yields for photochemical formation of hydrogen peroxide
in seawater. M.S. thesis, State Univ. of New York, College of
Environmental Science and Forestry.
MOBLEY, C. D. 1994. Light and water, radiative transfer in natural
waters. Academic.
NEALE, P. J., AND D. J. KIEBER. 2000. Assessing biological and
chemical effects of UV in the marine environment: Spectral
weighting functions, p. 61–83. In R. E. Hester and R. M. Harrison [eds.], Causes and environmental implications of increased UV-B radiation. The Royal Society of Chemistry.
NELSON, N. B., AND D. A. SIEGEL. 2002. Chromophoric DOM in
the open ocean, p. 547–578. In D. A. Hansell and C. A. Carlson [eds.], Biogeochemistry of marine dissolved organic matter. Academic.
, AND A. F. MICHAELS. 1998. Seasonal dynamics of
colored dissolved material in the Sargasso Sea. Deep-Sea Res.
45: 931–957.
POPE, R. M., AND E. S. FRY. 1997. Absorption spectrum (380–700
nm) of pure water. 2. Integrating cavity measurements. Appl.
Opt. 36: 8710–8723.
SHAW, G. E. 1983. Bio-controlled thermostasis involving the sulfur
cycle. Climate Chang. 5: 297–303.
SHOOTER, D., AND P. BRIMBLECOMBE. 1989. Dimethylsulphide oxidation in the ocean. Deep-Sea Res. 36: 577–585.
SIEGEL, D. A., AND A. F. MICHAELS. 1996. Quantification of nonalgal light attenuation in the Sargasso Sea: Implications for
biogeochemistry and remote sensing. Deep-Sea Res. 43: 321–
, AND OTHERS. 2001. Bio-optical modeling of primary production on regional scales: The Bermuda BioOptics project.
Deep-Sea Res. 48: 1865–1896.
LORENZI-KAYSER. 2002. Global distribution and dynamics of
colored dissolved and detrital organic materials. J. Geophys.
Res. 107, 3228, doi:10.1029/2001JC000965.
SIMÓ, R., AND C. PEDRÓS-ALIÓ. 1999a. Role of vertical mixing in
controlling the oceanic production of dimethyl sulphide. Nature
402: 396–399.
. 1999b. Short-term variability in the open
ocean cycle of dimethylsulfide. Glob. Biogeochem. Cycles 13:
STELFOX-WIDDICOMBE. 2002. Coupled dynamics of dimethylsulfoniopropionate and dimethylsulfide cycling and the microbial food web in surface waters of the North Atlantic. Limnol.
Oceanogr. 47: 53–61.
SLEZAK, D., A. BRUGGER, AND G. J. HERNDL. 2001. Impact of solar
radiation on the biological removal of dimethylsulfoniopropionate and dimethylsulfide in marine surface waters. Aquat.
Microb. Ecol. 25: 87–97.
A. F. MICHAELS, AND A. H. KNAP. 2001. Overview of the US
JGOFS Bermuda Atlantic Time-series Study (BATS): A decade-scale look at ocean biology and biogeochemistry. DeepSea Res. 48: 1405–1447.
An antioxidant function for DMSP and DMS in marine algae.
Nature 418: 317–320.
RUARDIJ. 1996. Model structure and analysis of dimethylsulfide (DMS) production in the southern North Sea, considering
phytoplankton dimethylsulphoniopropionate (DMSP) lyase and
eutrophication effects. Mar. Ecol. Prog. Ser. 145: 233–244.
1995. Photoproduction of carbonyl sulfide in south Pacific
Ocean waters as a function of irradiation wavelength. Geophys.
Res. Lett. 22: 215–218.
Microbial consumption and production of dimethyl sulfide
(DMS) in the Labrador Sea. Aquat. Microb. Ecol. 18: 197–
XIE, H., R. M. MOORE, AND W. L. MILLER. 1998. Photochemical
production of carbon disulphide in seawater. J. Geophys. Res.
103: 5635–5644.
YOCIS, B. H., D. J. KIEBER, AND K. MOPPER. 2000. Photochemical
production of hydrogen peroxide in Antarctic waters. DeepSea Res. 47: 1077–1099.
Received: 12 August 2002
Accepted: 19 January 2003
Amended: 5 February 2003
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF