Nov 20, 1994 - into the bottom of the bubbling chamber via a six-way valve. The He carrier gas ..... man and Cooper, 1988; Cooper and Saltzman, 1991; Bandy.
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 99, NO. Dll,
PAGES 22,819-22,829, NOVEMBER
20, 1994
Biogenic sulfur emissionsand aerosolsover the tropical South Atlantic
1. Dimethylsulfide in seawater and in the atmospheric boundary layer Tracey W. Andreae, Meinrat O. Andreae, and Gtinther Schebeske Max Planck Institute for Chemistry, Mainz, Germany
Abstract. We measureddimethylsulfide(DMS) in air (DMSa) and surface seawater (DMSw) on board the R/V Meteor during February-March 1991 on the tropical South Atlantic. Samplesfor the determinationof DMS a were taken througha fluorinated ethylene/propyleneTeflon inlet •33 m above sea level, preconcentratedby adsorption onto gold wool in quartz tubes, and analyzed by gas chromotographywith flame photometricdetection.The DMS a instrumentis fully automated,providingimproved precision,and processesup to four samplesper hour. Over most of the cruisetrack, which followedthe 19øSparallelbetweenBrazil and Africa, DMSw was significantlycorrelatedto climatologicallyaveragedchlorophyll concentrationsobtainedfrom coastalzone color scannerdata, suggestingthat remote sensingmay be useful for estimatingseawater DMS levels at least in some ocean regions. The cruise track proceededfrom waters of low productivity (off the coast of Brazil and in the subtropicalgyre) to higher productivity (the Benguela Current and the upwelling region off Namibia and Angola); meteorological conditionswere steady with consistenteasterly winds. DMS values for air and water were
low(•50 pptand1-2nmolL -i , respectively) in theareasof lowproductivity and increased simultaneously (• 100-300 pptand3-15nmolL-i) asproductivity increased. DMSsea-to-air fluxes(average 7.3/•molm-2 d-1) werecalculated based ondifferent parameterizations;for the study region the differencesbetween the results obtainedfrom the differentmodelswere minor. DMSa was stronglycorrelatedto its emissionflux from the sea surfaceas estimatedfrom DMSw and meteorologicalparameters.This suggests that the air/seatransferparameterizationsused are suitablefor providingestimatesof DMS flux from the oceans.
fractionof totalbiogenic sulfuremissions (65-125Tg S yr -•)
Introduction
Because its atmospheric photooxidation is the main source of cloud condensation nuclei (CCN) in the remote marine atmosphere, dimethylsulfide (DMS) plays an important role in global climate regulation [Charlson et al., 1987]. DMS and its precursors are produced in seawater by phytoplankton, and after crossing the air-sea interface, DMS is present in the atmosphere. Here it is oxidized into sulfate particles, which act as CCN and thus affect the reflectance of clouds and therefore the Earth's radiation budget and climate. The Charlson-Lovelock-Andreae-Warren (CLAW) hypothesis[Charlson et al., 1987]proposesthat phytoplankton vary their production of DMS in responseto changesin surface temperature and sunlight and modulate the planet's climate in a feedback loop. Similar feedback mechanisms have been proposed by Nguyen et al. [1983] and Meszaros [19881.
Although the precise feedback mechanisms in the biogeochemicalsulfur cycle are yet to be understood, it is now clear that DMS is the predominant sulfur gas emitted from
[Andreae and Jaeschke, 1992]. DMS is emitted by phytoplankton both directly [Lovelock et al., 1972;Andreae et al., 1983; Vairavamurthy et al., 1985] and as a breakdown product of dimethylsulfonium propionate (DMSP) [Challenger and Simpson, 1948;Belviso et al., 1990; Kiene, 1992], which is produced by marine algae for the purpose of osmoregulation [Vairavamurthy et al., 1985; Dickson and Kirst, 1987a, b]. DMS is released in larger quantities when cells are under grazing pressureby zooplankton [Dacey and Blough, 1987], during senescence [Nguyen et al., 1988], or during bacterial putrefaction after a bloom [Bremner and Steele, 1978]. Global estimates of DMS flux differ by a factor of 2 to 3 between different investigators. A global marine DMS emis-
sionof •39 Tg S yr-• was estimatedby Andreaeand Raemdonck [1983]; this estimate did not change significantly
in a recentreevaluation (35 - 15Tg S yr -• [Andreae1990]). A considerably lower value, •16(8-32) Tg S yr -• was
proposed by Bates et al. [1987b]. Using a relationship between DMS flux and incident sunlight proposed by Bates theoceans(16-50Tg S yr-1) [Bateset al., 1987b; Andreae, et al. [1987a] and a general circulation model to estimate 1990; Erickson et al., 1990]. This accounts for a large transfer velocities, Erickson et al. [1990] obtained an estiCopyright 1994 by the American Geophysical Union.
mateof 15 Tg S yr -• for the globalmarineDMS source.
Paper number 94JD01837.
Even though the ranges of uncertainty given by these authors overlap, the fact that this disparity has remained
0148-0227/94/94 JD-01837505.00
22,819
22,820
ANDREAE
ET AL.: DMS IN AND OVER THE TROPICAL
SOUTH
ATLANTIC
dimensional simulation of sulfur cycling in the marine boundary layer, submitted to Journal of Geophysical ReDMS database. search, 1994; hereinafter referred to as paper 2.) An invesThe potential reasons for the prevailing uncertainty in the tigation of the relationship between atmospheric DMS and globalDMS flux are the uncertaintyin the transfervelocity atmospheric aerosol particles, including CCN, will be preterm in the flux equation and differences in the databasesand sented by M. O. Andreae et al. (Biogenic sulfur emissions extrapolation approachesused. However, in spite of persist- and aerosols over the tropical South Atlantic, 3, DMS, ing uncertainty about the absolute value of the transfer aerosols, and cloud condensation nuclei, submitted to Jourvelocity as a function of wind speed, the globally averaged nal of GeophysicalResearch, 1994; hereinafter referred to as transfer velocities used by Bates et al. [1987b], Erickson et paper 3.)
unchanged since 1987 points to a need for improvement in our techniques for obtaining flux estimates from the existing
al., [1990], and Andreae [1990] all fall within the narrow
rangeof 10-13cm h-1. Thereforeit doesnot appearthat differencesin estimated transfer velocity can account for the different flux estimates by these authors. These differences must then result from the different
surface seawater DMS
concentrations used and from differences in extrapolation approaches. Bates et al. [1987b] use data sets from the Pacific Ocean; those used by Andreae [1990] are of worldwide origin but with an emphasis in the Atlantic Ocean. Bates et al. used a geographical extrapolation based on latitude
and assumed that the latitudinal
distribution
in the
Pacific applies to all oceans, Erickson et al. related surface ocean solar radiation to DMS flux, and Andreae extrapolated by biogeographical regions. The resulting average surface
Methods
Sampling Surface water samplesfor the determination of dissolved DMS were taken every 4 hours from a continuous-flow seawater pump with an intake depth of approximately 5 m. Surface samplesfrom hydrocasts or from a hand-held bucket were also collected, and no significant difference between collection methods was observed. Samples were drawn into linear polyethylene bottles, and sample degassing during transfer was avoided. Whenever possible, samples were analyzed immediately. Otherwise, samples were stored in the dark at near in situ temperatures (samplesstored in this
seawaterDMS concentrations rangefrom 1.6 nmol L -1 manner are stable for 24-48 hours [M. O. Andreae and (Ericksonet al.) through1.8 nmolL -1 (Bateset al.) to 3.0 Barnard, 1983]. No samples were stored longer than 10 nmolL -1 (Andreae),whichaccounts for the differences in hours. these authors'
flux estimates.
al., 1990]. Here we present a 6-week continuous data set for atmospheric and seawater DMS taken along 19øS from South
Air samples were taken through an inlet secured to the mast at approximately 33 m above sea level. The inlet was protected from moisture by a plastic funnel; a Teflon cyclone at the inlet prevented sea salt aerosol from entering the sample line. Approximately 10 m of 9.5-mm-OD Teflon fiuorinated ethylene/propylene (FEP) tubing connected the inlet to the DMS analyzer. A plug of cotton held in a 47-mm-diameter Teflon filter housing was installed in the intake line just before the analyzer to eliminate negative interference by ozone and other oxidants in the DMS determination [T. W. Andreae et al., 1993]. The DMS analyzer used is fully automated and can process up to four samples per hour. For routine DMS measurements a sample collection time of 30 min was selected. Two samplesper hour were collected and immediately analyzed continuously throughout the cruise, with the exception of daily calibration periods. Sampling was automatically interrupted by a Weathertronics sampler controller if the relative wind direction was greater than 90ø off the bow. This was rarely necessary, however, since the ship's course was into the prevailing easterly trade winds during most of the cruise. Meteorological information was acquired by the ship's on-board system, which is operated by the German Meteorological Service. The data recorded include temperature, humidity, pressure, wind speed and direction, seawater temperatureand salinity, global radiation, UV radiation, and two daily radiosonde soundings.
America to Africa. The relationships between marine productivity (as reflected by satellite-derived chlorophyll) and
Analysis
Thus if we want to answer the question "How much DMS is emitted from the oceans, from where, and when?," we have to make sure we have a representativedatabaseon the distribution
of DMS
in the surface
ocean
and that we are
applying proper methodsfor global extrapolation. Collection of more data, especially with simultaneousmeasurementsof biological and chemical reference variables and from regions and times of year where data gaps currently exist, is still important to improve the present DMS database. The more ditficult problem, however, appears to be the selection of a rational techniquefor extrapolation. Attempts to relate DMS concentration to parameters such as phytoplankton biomass, nutrient limitations, chlorophyll a (chl a), seawater temperature, salinity, and incident solar radiation have not revealed any simple relationships [Andreae and Barnard, 1984; Barnard et al., 1984; Bates and Cline, 1985; Bates et al., 1987b; Holligan et al., 1987; Turner et al., 1988;Iverson et al., 1989; Andreae, 1990; Bt2rgermeisteret al., 1990; Leck et al., 1990]. The use of satellite imagery to obtain estimates of DMS production is receiving a lot of interest; however, determining the best parametersto use is ditficult. Investigationshave included coccolith reflectance and chl a, among others [Malin et al., 1992; Matrai and Keller, 1993; Thompsonet
seawater
DMS
concentrations
and between
fluxes calculated
from seawater concentrations, meteorological data, and atmospheric DMS levels will be discussed.A detailed discussion and model of the sulfur cycle in the atmospheric boundary layer during Meteor 15/3 is given in the forthcoming paper by K. Suhre et al. (Biogenic sulfur emissionsand aerosols over the tropical South Atlantic, 2, One-
Seawater samples were analyzed for DMS using the method of Andreae and Barnard [1983, 1984]. Samples are drawn into the analytical system through a 13-mm "extra thick" glass fiber filter (Gelman Sciences) to remove plankton and biological debris. Omissionof this step resultsin the release of DMS from particulate material during the gasstripping stage, causing DMS levels which may be double
ANDREAE ET AL.: DMS IN AND OVER THE TROPICAL SOUTH ATLANTIC
22,821
Sample Inlet Calibration
•
Stream
Inlet Cotton Scrubber
FPD
Amplifier I
./ '
i
L-1
i
I
i iI
r- 'J
I i
I I
i
I
i
i
Ii
I
I
I
I
I
I
I
I
I
L.....
I
u
I
I
' ^•1
I
I
i
I
I
,
' --Iam? i /,N,Ii Oo,u Ii
Controller
'
IJacketed I
I
I
i
t
I ea' I
Interface
I PC with E-lab
Printer
:::::::::::::::::::::::::::: N2 :!:i:i: I
_J
:::::::::::::::::::::::::::::::::::::::::::
Figure 1. Schematic representationof the automated analyzing system for DMS in air. Electrical connectionsare shownas dotted lines, gas flows as full lines.
those from filtered samples.The sample is drawn into a calibratedloop of 3.2-mm-ODTeflon tubing.After the loop is flushed,the samplealiquot is injectedthrougha glassfrit into the bottomof the bubblingchambervia a six-wayvalve. The He carriergasis passedthrougha scrubberpackedwith activated charcoal and molecular sieve. The sampleis degassedfor 10-20 min dependingon the sample volume, which canrangefrom 1 to 10mL (dependingon the expected concentrationsin the seawater being analyzed). From the bubblingchamberthe carrier gas stream passesthrougha dryingtube filled with granularK2CO3 (Alfa Products)and
into a U-shapedtubefilled,withchromatographic packing (15% OV3 on ChromosorbW-AW-DMCS). During the degassingperiod the tube is immersedin liquid nitrogen.After the strippingtime the liquid nitrogenis removedand the tube is heated, actingas a simple,temperature-programmed gas chromatograph.In most cases, DMS is the only volatile sulfur compounddetectableat the samplevolumesused. A specially designed, sulfur-selective, highly sensitive flame photometric detector is used to detect DMS; its detection limit is 30 pg S (DMS). Calibration was by addition of gravimetrically prepared standards of DMS in ethylene glycol to degassedseawater samples. The results are recorded automatically using a chromatographysystem (ELab, OMS Tech, Miami, Florida) installed in a Toshiba laptop computer. Precisionis 3%, and accuracy is better than 5%.
DMS in air was determinedusingan automatedsampling and analysissystem(Figure 1) basedon the techniqueof M. O. Andreae et al. [1985] and T. W. Andreae et al. [1993]. The systemhastwo samplingchannels,oneof whichis collecting a samplefrom the air streampassingthroughit, while DMS is being desorbedand analyzed from the other. At the end of the samplingintervalthe eight-portvalve is actuated,andthe sampling channels switch roles. DMS is collected from the
sampleair streamby chemisorptionto gold wool in a quartz tube. For desorption the quartz tube is heated to --•300øC while a H2 stream passesthrough it. This gas stream then passesthrough a 30-cm-long,U-shaped glasstube (6.2-mm OD) filled with chromatographicpacking (15% OV3 on
ChromosorbW-AW-DMCS). During desorptionthis tube is cooledto --•- 100øCby blowingthe vapor from boilingliquid nitrogenthrough an aluminumjacket surroundingthe tube. When desorptionis completeafter 3 min, the voltage suppliedto the resistorin the liquidN2 reservoiris switchedoff, and power is applied to the heating wire coiled around the U-shaped tube. This causes the gases, including DMS, trapped on the tube to elute into the detector, which is identical in design to the one used for the determination of DMS in seawater(see above). The resultingchromatogram is registeredin a data acquisitionsystem(E-Lab, OMS Tech, Miami, Florida), which also controlsthe timing of the steps of the samplingand analysissequence.Following the analysis step (3 min) the gold tube is heated for 5 min to about 500øC in the H 2 stream to regenerate the gold surface. Heating is then discontinued,and the trap is cooled to room temperatureand flushedwith H 2 until the samplingperiod for the other channel is complete. The systemis calibratedby addingDMS from a permeation device through a two-step gas dilution system to a stream of purified air. By adjusting the dilution volume, concentrationsbracketingthe atmosphericlevels (typically 30 and 200 parts per trillion (ppt)) are prepared. This calibration gas stream is switched into the inlet of the
analyticalsystem,aheadof the cotton scrubber.The system is calibrated by analyzing duplicate samples of the two standardson each channel at least once daily. Calibration drift is usually minor, about 5-10% per day, and is corrected for in the data processing.The precisionof the systemis 5% or better; the accuracy of a manual version of this system
22,822
ANDREAE
ET AL.: DMS IN AND OVER THE TROPICAL
SOUTH ATLANTIC
.:.
'-0,04 0..':'22 -1.:'26 '7.0-"8.'39.-,..B :mg/m' Figure 2. The cruisetrackof Meteor 15/3superimposed on the distributionof chlorophylla in the South Atlantic. The chlorophylldistributionrepresentsthe mean of all chlorophyllmeasurementsmade by the coastal zone color scanner over the period 1978-1986.
was establishedduring a blind intercalibration exerciseto be
knots(7.4 km h-1) makingpossiblethe collectionof data
6% iT. W. Andreaeet al., 1993].
with high spatial resolution. The cruise track cut across a large regional gradient of productivity related to the differencesbetween the oligotro-
Results
and Discussion
DMS and Chlorophyll a in Surface Seawater
The cruise track for sampling aboard the R/V Meteor (cruise 15/3) in the South Atlantic is shown in Figure 2, superimposed on a map of satellite-derived chlorophyll concentration. The cruise began at Vitoria, Brazil, followed the nineteenthparallel acrossthe South Atlantic, and ended at Pointe Noire, Congo(February 10 to March 23, 1991).The ship remained within the southeasterlytrade wind circulation for the duration of the cruise; a constant regime with the ship heading into the wind eliminated the possibility of contamination from the ship for most of the cruise. Clean marine air, with only occasionallow levels of atmospheric pollution, and steady meteorological conditions characterized the cruise. Due to the extensive station work being performed as part of the World Ocean Circulation Experiment the ship progressedat an average speed of about 4
phic waters of the Brazil Current and the South Atlantic Gyre and the upwelling of the Benguela Current system (Figure 3). These water massescan be clearly identified on the basis of their temperaturesand the distribution of chlorophyll. The chlorophyll data shown in Figures 2 and 3 were obtained from the coastal zone color scanner (CZCS) flown on Nimbus 7 during 1978-1986 (in situ chlorophyll data were not available for this cruise). The values in Figure 3 represent the annual mean chl a values for the pixels along the cruisetrack, plotted alongthe sametime axis as the seawater
temperaturesmeasuredfrom the ship. Higher chl a values are evident in the shelf waters along the Brazilian coast, followed by very oligotrophic, warm waters of the South Atlantic Gyre. Transition into the BenguelaCurrent regime occurred around 6øW (March 4-5); the temperature record showsno distinct break, but the salinity data as well as the DMS concentrations(discussedbelow) show a clear discon-
ANDREAE
ET AL.' DMS IN AND OVER THE TROPICAL
10
301 28
•
26
E
24
•
22
SOUTH ATLANTIC
22,823
15
•er temperature 1•
0.1 •
hyll 18 ...... 1O-Feb
510
' ' ' ' ' ' ' i , , .... 20- Feb 02-Mar
, ...... 12-Mar
5
• 0.01 22- Mar
0 0.02
1991
0.1
1.0
Chlorophyll a (monthly average),mgm-3
Figure 3. Chlorophyll a and surface seawater temperature along the cruise track (plotted versus time). The chlorophyll distribution represents the mean of all chlorophyll measurements made by the coastal zone color scanner over the period 1978-1986.
FiBare 5, D•S concentration in su•ace seawater versus chlorophyll •. The chlorophyll data represent monthly mean values collected dudn8 February and •arch over the period ]97•]986. Feb•ary data are shown for the pa• of the cruise that fell into Feb•ary; •arch data are shown for the •arch
tinuity in this region. The coldest waters were found in the upwelling region off Namibia around March 18, coinciding with high climatological chl a values. Warmer waters were encountered again on the northbound transect toward Pointe Noire, Congo. Seawater DMS concentrations are shown together with CZCS chlorophyll in Figure 4. Since no CZCS data are available for the actual time of the cruise, mean values for February (1978-1986) are used for the part of the cruise falling into that month, and mean March CZCS data are shown for the March segment of the cruise. DMS concentrations in the Brazil Current were approximately 2 nmol L -• at the start of the cruise then decreasedto values near
section of the cruise.
approximately 44 nmolL- • beingrecorded onMarch17,in the Benguela upwelling region off Namibia. Comparison between the chl a data and the DMS
concentration
shows a
reasonable agreement in the large-scale trend but also some clear differences,for example, in the location of the peak off Namibia. The climatological chl a data agree well with the observed temperature distribution, predicting the highest plankton levels in the coldest waters of the upwelling area. The DMS maximum, on the other hand, occursjust outside of the maximum upwelling and its phytoplankton bloom. This behavior has been observed previously in the Peru upwelling region [Andreae, 1985]. It probably reflects the fact that DMS and its precursor, DMSP, are produced only 1 nmolL -• andremained closeto thisvaluethroughout the South Atlantic Central Gyre. Upon entering the Benguela in small quantities by the diatoms, which predominate in the Current region, around March 4, we observed an increase to first, inshore stagesof the upwelling bloom. The phytoplankabout4 nmolL-1. Thelevelsgradually increased aswedrew ton organismswhich are more prolific in their DMS produccloser to the coast of Africa, with the highest values of tion are typically found in the later, offshore stages of the upwelling bloom, coinciding with the position of the DMS maximum. A plot of DMS versus CZCS chlorophyll (Figure 5) reflects the same relationships: for the range of chloro15
•
10
r•
5
1.0
phylllevelsup to about0.25 mg m-3 a clearcorrelation exists between the two parameters. At higher chlorophyll levels this correlation breaks down, presumably as a result of the transition into a different plankton ecological regime. A rigorous statistical analysis of this data set is problematic, since DMS measurements taken at specific times in 1991 must be compared with annual or seasonal average chlorophyll data collected over a time span of 8 years and several years before the cruise. If the 783 hourly mean DMS values from Meteor 15/3 are regressedon the annual average
chlorophyll data,a relativelypoorr 2 of 0.20 is obtained, ........................................
10-Feb
17-Feb
24-Feb
0.01
03-Mar
10-Mar
17-Mar
1991
which is not much improved (to 0.25) by deletion of the 11 data points correspondingto the sharp DMS maximum near the Benguela upwelling. Regression against the log-
transformed chlorophyll datagivesa muchbetterr 2 of 0.52, andafterremovalof the 11 outliers,r 2 = 0.58 (Figure5).
Figure 4. DMS concentrations in surface seawater and chlorophyll a along the cruise track (shown as time series plot). The chlorophyll data represent monthly mean values collected during February and March over the period 19781986. February data are shown for the part of the cruise that fell into February; March data are shown for the March
Using the CZCS data from February and March for the correspondingcruise sectionsyields a slightly better overall
section of the cruise.
transformation does not further improve this fit. These
fit (r 2 = 0.30), andwhenonlythe datawith chl a values
