Air-Sea Exchange Of CO2: A Multi-Technology Approach

Journal of Marine Systems 86 (2011) 1–9

Contents lists available at ScienceDirect

Journal of Marine Systems

j ourna l homepage: www.e lsev ie r.com/ locate / jmarsys

Observed carbon dioxide and oxygen dynamics in a Baltic Sea coastal region

Karin Wesslander a,⁎, Per Hall b, Sofia Hjalmarsson b, Dominique Lefevre c, Anders Omstedt a,Anna Rutgersson d, Erik Sahlée d, Anders Tengberg b

a Department of Earth Sciences, University of Gothenburg, Box 460, SE-405 30 Gothenburg, Swedenb Department of Chemistry, University of Gothenburg, SE-412 96 Gothenburg, Swedenc Aix-Marseille Universite´, CNRS, LMGEM-UMR 6117, Laboratoire de Microbiologie, Géochimie et Ecologie Marines, Campus de Luminy-Case 901, 13288 Marseille cedex 9, Franced Department of Earth Sciences, Air- and Water Science, Uppsala University, Uppsala, Sweden

⁎ Corresponding author.E-mail address: [email protected] (K. Wes

0924-7963/$ – see front matter © 2011 Elsevier B.V. Aldoi:10.1016/j.jmarsys.2011.01.001

a b s t r a c t

Article history:Received 8 December 2009Received in revised form 25 August 2010Accepted 18 January 2011Available online 28 January 2011

Keywords:Baltic SeaCarbon dioxideOxygenpCO2

Air–sea exchange

In April 2006, diurnal variations of carbon dioxide (CO2) and oxygen (O2) in the surface water east of Gotlandin the Baltic Sea were investigated with a unique multitechnology approach. Several parameters weremeasured simultaneously providing an overview of the CO2 system.Moored instruments were continuouslyrecording partial pressure of CO2 in the surfacewater (pCO2

w), currents, mixing, waves, salinity, temperatureand O2. Measurements of total alkalinity (AT) and dissolved inorganic carbon (CT) were taken from R/VSkagerak. These measurements were converted to pCO2

w to support the continuous pCO2w data and also

calculate the air–sea exchange of CO2. Additionally, the time derivatives of O2 and CT concentrations inthe water were determined using incubations and a Productivity Autosampler (PA). O2 and pCO2

w weresignificantly anti-correlated and periods dominated of either biological processes, mixing, air–seaexchange or a combination of these were detected. O2 and pCO2

w had a daily cycle and variations occurredon the 1 h time scale. In April 2006, the seawas a CO2 sink and the averaged parameterized air–sea exchangewas −1.0±0.6 mmol m−2 h−1.

slander).

l rights reserved.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Increasing atmospheric levels of CO2 is affecting both theatmosphere, through increased concentrations of green house gases,and the sea, through interaction with the water acid-base (pH)balance. When a new state of equilibrium is set up due to higher CO2

levels in the atmosphere, the partial pressure of CO2 in the sea (pCO2w)

is also expected to increase. Measurements and estimates of the air–sea exchange of CO2 show that the sea is acting both as a sink and asource for atmospheric CO2 depending on the location and character-istics of the region as well as on the season (e.g. Bozec et al., 2006).Continental shelves are, during present climate conditions, estimatedto be significant sinks for atmospheric CO2 whilst estuaries function assignificant sources (Borges et al., 2006). The importance of the coastalzones has previously been pointed out through its considerable role inthe biogeochemical cycles (Gattuso et al., 1998; Bozec et al., 2005),one reason being the large input of terrestrial organic matter andnutrients. A recent model study (Omstedt et al., 2009) indicates thatthe Baltic Sea was a source for atmospheric CO2 before theindustrialization. During modern industrialization time with eutro-phication, the Baltic Sea has however turned into a system that actsboth as a source and a sink. The seasonal and inter-annual variations

of the carbon cycle in the central Baltic Sea illustrates large variabilitybut as a mean over the 1994–2008 period this region acted as a sourceof atmospheric CO2 (Wesslander et al., 2010).

The marine CO2 system involves several pathways such as air–seainteraction, mixing, and biological processes. To understand thesystem there is a need for multi-parameter in situ measurements withhigh temporal resolution. Preferably, data of all the different processesshould be collected at the same time and at the correct samplingfrequency to capture the major variability. The variability in pCO2

w andair–sea exchange of CO2 is often interpreted from monthly measure-ments. However, it is important to resolve the variations on shorttimescale in order to understand the variations between and withinthe monthly means. Short timescale variations can be caused by e.g.biological processes, upwelling events, river inflow, advection, andmixing processes.

The CO2 system in the Baltic Sea has been studiedmore intensivelyfor the last 10 years, but the temporal resolution has often been coarsewith typical weekly to monthly measurements. Several studies on themarine CO2 system in the Baltic Sea have been performed in the opensea, mainly in the eastern Gotland Basin, an area representing themean Baltic Sea conditions (Winsor et al., 2001, 2003). Thomas andSchneider (1999) concluded that the CO2 system in the Baltic Sea isstrongly influenced by lower salinity in the surface water duringspring and summer and by higher salinity in the surface water duringautumn and winter. Schneider et al. (2003) found that during thebiologically productive season, the consumption of CO2 is the major

2 K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

control of pCO2w changes. A recent study of the ocean pH dynamics has

highlighted the urgent need for more spatially distributed andtemporally intensive studies (Wootton et al., 2008). Studies on theCO2 system in the Baltic Sea with high temporal resolution is e.g. Kusset al. (2006) who carried out an hourly measurements time series ofpCO2

w and O2 from a platform in the Arkona Basin. They concluded thatto better understand the CO2/O2 system, measurements should bedone on short timescales rather than monthly cruises which mostlikely do not measure the low pCO2

w levels recorded in late summer.There is an advantage in combiningmarine studies of CO2 and O2 sincethey are closely biogeochemically related through photosynthesis andrespiration, but have different dynamics in the water column. Someinvestigations have been carried out with automatic measurements ofpCO2

w and O2 using pumped systems on cargo ships (Schneider et al.,2006, 2007) providing surface water data over certain ship routeswith a high resolution in both time and space. A recent study thatcouples atmospheric measurements with processes in the water isRutgersson et al. (2008) who used buoy measurements of pCO2

w andatmospheric measurements of the air–sea exchange from a fieldstation. Rutgersson et al. (2008) concluded that there are significantdifferences between pCO2

w near land and pCO2w at open sea mainly

during intense biological activity and upwelling events. Investigationswith high temporal resolution on the CO2 system from other coastalregions include e.g. Borges and Frankignoulle (1999). They studieddaily variations on pCO2

w along the Belgian and southern Dutch coastalareas and found that daily variations in pCO2

w were, in that region,controlled by the tidal and biological cycle.

The main purpose of the present investigation was to, in a coastalregion in the central Baltic Sea, combine the physical, chemical, andbiological processes involved in the marine CO2 system, with a hightemporal resolution and with as complete instrumentation aspossible. By using a multi-technology approach we were aiming atbeing able to distinguish and quantify the different processes whichaffect the balance and exchange of CO2 and O2 in the water. This is apioneer study for the feasibility to tackle these objectives in order tohave a sound strategy for a seasonal survey.

2. Materials and methods

2.1. Outline of the study

This work is based on measurements from the coastal region offthe east coast of Gotland, and from the small flat island Östergarn-sholm (57° 25 N, 18° 59E), 4 km east of Gotland (Fig. 1). Wemeasuredatmospheric parameters from a 30 m tall tower situated onÖstergarnsholm that has been operational for several years (e.g.Rutgersson et al., 2001, 2008). The field station is run and owned byUppsala University. For a complete description of the tower see e.g.Högström et al. (2008). Simultaneously with the atmosphericmeasurements sea parameters were collected, just outside the island,with a unique combination of in situ measurements. The pCO2

w,temperature, and incoming light (photosynthetically available radi-ation: PAR)were directlymeasured at 5 m below the surface and dailyprimary production/respiration measurements were performed au-tomatically and manually. At 5 m and at four additional depthsdissolved oxygen concentration data (O2) were collected everyminute. To obtain information on water circulation and waves atthis 22 m deep site a high resolution current profiler was placed onthe bottom. In addition salinity and temperature were continuouslymeasured at the surface and the bottom. The quality of the in situmeasurements was verified by frequent ship measurements (R/VSkagerak), which also gave additional information on changes ofnutrients in the water column.

The measurements were done between April 5 and 10 2006.During this time of year, the Baltic Sea surface water is typically colderthan the atmosphere, which gives a stable stratified atmospheric

boundary layer effectively decreasing atmospheric turbulence. It ispreviously known that the presence of stable stratification acts like alid on the surface effectively limiting air–sea exchange of heat andhumidity (Rutgersson et al., 2001), implying also a reduction in theair–sea gas exchange.

To expand the study of the marine CO2 system we also calculatedthe air–sea gas exchange of CO2 using three different parameteriza-tions; Wanninkhof (1992), Wanninkhof and McGillis (1999) andWeiss et al. (2007).

2.2. Meteorological measurements in the tower

2.2.1. Temperature, relative humidity, and windThe instrumentation consists of cup anemometers and wind vanes

at 5 levels for measurements of wind speed and wind direction (7,11.5, 14, 20 and 28 m above the tower base). Ventilated and radiationshielded thermocouples were mounted at the same levels fortemperature measurements. Relative humidity was measured at 8 mabove the tower base. These measurements had a sampling frequencyof 1 Hz.

2.2.2. Partial pressure of CO2 in the atmosphereThe average atmospheric CO2 molar fraction was measured once

every half hour with an infra-red gas analyzer system at a height of9 m (PP-systems). A molar fraction of dry air, χc, was calculated usingmeasured humidity. χc was converted to the partial pressure of CO2 inthe atmosphere (pCO2

a ) through:

pCOa2 = χcpa ð1Þ

where pa is the measured air pressure (hPa) corrected for humidity,i.e. where the partial pressure of the water vapour, e, has beenremoved. Partial pressure of water vapour is calculated from themeasurements of relative humidity (RH) and temperature (T) in thetower:

e = RH·es Tð Þ ð2Þ

where es is the saturation water vapour pressure, a function of the airtemperature.

The zero value was set using N2-gas. The span was set with a388 ppm CO2 in dry air mixture.

2.3. Sea water measurements

2.3.1. Dissolved oxygen, water circulation, salinity, and temperatureTo study the coupling between the variations in CO2 and O2

concentrations, a 12 channel data logger (DL7, http://www.aadi.no),that is otherwise used to log sensors on an autonomous benthic lander(e.g. Tengberg et al., 2003; Ståhl et al., 2004; Brunnegård et al., 2004),was suspended from an anchored buoy about 1 m below the surfaceand set to collect data at 1-minute intervals. O2 was measured at 4water levels (0.6, 1.4, 1.9 and 4.8 m below the surface) with stable andaccurate oxygen optodes (Tengberg et al., 2006). The absoluteaccuracy of these sensors was tuned for these measurements withan automated high precision Winkler titration system linked to aphotometric end point detector (Williams and Jenkinson, 1982) toapproximately ±1%. Possible sensor drift was verified, by watersampling and subsequent Winkler titration, before, during, and afterthe deployments and was not detected. In addition to oxygen:temperature, salinity, particles (turbidity), and pressure were logged1 m below the surface.

To get detailed information on water circulation a RecordingDoppler Current Profiler (RDCP 600, http://www.aadi.no) wasdeployed on the bottom during the field-work. The RDCP was set tomeasure current speed and direction both vertically and horizontally

Fig. 1. The Baltic Sea, Gotland, and Östergarnsholm. The island Östergarnsholm is marked with a black circle.

3K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

every 2 s in 20 cm sections from the bottom to the surface. Apart fromthe current measurements the instrument was equipped with sensorsto record turbidity, pressure (giving water level and wave informa-tion), salinity, temperature and oxygen concentrations, with optode,in the bottom water (approximately 1 m above the seafloor). Thesesensors were logged at 10-minute intervals.

The sea level difference across themeasuring sitewas generated frompermanent sea level measurements at Landsort, north of, and Kung-sholmsfort, south ofGotland. Thesedatawere available from the SwedishMeteorological and Hydrological Institute (http://www.smhi.se).

2.3.2. Partial pressure of CO2 in the waterThe pCO2 in the water was measured in two ways. Firstly, a SAMI

instrument (Submersible Autonomous Moored Instrument for CO2

from http://www.sunburstsensors.com) was moored at 5 m depthgiving hourly information about pCO2

w andwater temperature. Briefly,the principle of operation is to measure the optical absorbance of a pHindicator solution that is equilibrated with sea water pCO2 via a gas-permeable membrane. The accuracy of the instrument was ±3 μatmaccording to the manufacturer and previous experiments show nodrift for several months of measurements (e.g. DeGrandpre et al.,1995, 1999, 2000). The SAMI hourly measurements represent a time-weighted average over the last minutes of each hour. Secondly, pCO2

w

was calculated from measurements of total dissolved inorganiccarbon (CT) and total alkalinity (AT) from water samples taken fromR/V Skagerak. Coulometric titration was used to determine CT(Johnson et al., 1987), and AT was measured by potentiometrictitration according to Haraldsson et al. (1997). The precision wasestimated by calculating the standard deviation for duplicate samples,giving ±1 μmol L−1 for AT and ±2 μmol L−1 for CT. The accuracy of CTand AT was set by analyzing Certified Reference Material (CRM)supplied by A. Dickson, Scripps Institution of Oceanography (USA).From the measured AT and CT we calculated the pCO2

w using thesoftware CO2SYS (Lewis and Wallace, 1998). This method is reliable

and we used it to support the SAMI data. The uncertainty in thecomputed pCO2

w was estimated to approximately ±15 μatm.

2.3.3. Productivity autosamplerThe productivity autosampler system (PA) estimated the biological

contribution to the CO2 fluxes within the surface water by measuringchanges in O2 during day and night. In situ water sampling andincubation were automated and a pulsed oxygen sensor (Langdon,1984)measured the concentration of O2within the incubation chamberevery 10 min. The sensor consumed about 2 nmol O2 per sample. Overthe course of 24 h the sensor would consume only 0.02% of the O2

content of the 2-dm3 incubator. Time course of temperature andphotosynthetically available radiation (4 pi PAR, Biospherical Instru-ments QSP200) within the incubation chamber were also obtained. Seefull description in Langdon et al. (1995). The accuracy of the O2 sensorwas ±3 μmol O2L−1 and the precision ±0.1 μmol O2L−1. The accuracyof the PAR sensor was about 5% and the precision was ±1 μE m−2 s−1.

The chamber, which was submerged at 5 mwater depth, was open1 h before dawn and remained open for 1 h before closing for thefollowing 23 h. The dusk to dawn oxygen signal was representativefor the changes during night time whereas the dawn to dusk signalwas representative for the day time.

2.3.4. Bottle incubationsChanges in the concentration of O2 and CT over 24 h bottle

incubations (BI) were measured in situ during day and night,respectively. The rates were measured at 5 m depth at four occasions.Three sets of four replicates were collected in 125 cm3-borosilicateglass bottles. One set of samples was fixed immediately to measurethe O2 and CT concentrations at time zero; the second set wasincubated for 24 h in the dark and the remaining set was incubatedunder in situ light conditions. Dissolved O2 concentration wasmeasured using an automated high precisionWinkler titration systemlinked to a photometric end point detector (Williams and Jenkinson,

4 K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

1982). CT concentration was analyzed with the same method asdescribed in the section Partial pressure of CO2 in the water.

Pooled standard deviations for the O2 titration were 0.24, 0.32 and0.36 μmol O2L−1 for time zero, 24 h dark and 24 h light incubations,respectively. Pooled standard deviations for CT titrations were 1.5, 1.2and 1.3 μmolCTL−1 for time zero, 24 h dark, and 24 h light incubations,respectively. The overall coefficient of variation for O2 and CT titrationswas 0.06%. Day time rates of changes in O2 and CT were calculated as thedifference in the O2 or CT concentration between “light” incubatedsamples and “time zero” samples. Night rates were calculated as thedifference between “dark” incubated samples and “time zero” samples.Night rates are expressed as a negativeO2 rate and a positiveCT rate. Theprecision of the mean O2 rates was ±0.2 μmol O2 L−1 d−1, and theprecision of the mean CT rates was ±0.9 μmolCTL−1 d−1 as estimatedusing the standard error from the analyses of quadruple samples sets.

2.3.5. NutrientsThe concentration of phosphate and nitrate was analyzed fromwater

samples taken by a cable operatedCTD/Rosett system fromR/VSkagerak.Phosphatewas analyzed according to Carlberg (1972),Murphy and Riley(1962), and Strickland and Parsons (1968). Nitrate was analyzedaccording to Carlberg (1972), Nordforsk Miljövårdssekretariatet(1973), and Svensk Standard, SS028133 (1991). Precision was deter-mined from duplicate samples and was for phosphate ±0.05 μmol L−1

and ±0.07 μmol L−1 for nitrate.

a

b

c

d

5 6 7 8 9 10 5

Fig. 2. Physical parameters. a)Wind direction (line) and horizontal current direction at 5 m (at the tower. c) Sea level difference between Landsort, north of, and Kungsholmsfort sod) Horizontal current speed at 5 m. e) Significant wave height. f) Density at 1 m (black) and dh) Photosynthetically available radiation (PAR) at 5 m depth.

2.4. Calculated air–sea exchange of CO2

The air–sea exchange of CO2 (FCO2) was parameterized using thebulk formula:

FCO2= K0k pCOw

2 −pCOa2

� � ð3Þ

where K0 is the salinity and temperature dependent solubilityconstant, which was calculated using the empirical formulationsin Weiss (1974) and k is an exchange coefficient, the gas transfervelocity.

The success of the bulk approach is dependent on a correct value ofk. Several attempts of parameterizations have been published (e.g. seethe review in Liss et al., 2004) with the goal to express k as a functionof wind speed. The three parameterizations of Wanninkhof (1992),Wanninkhof andMcGillis (1999) andWeiss et al. (2007) were used inthis study for comparison. Wanninkhof (1992) presented thefollowing parameterization:

kW92 = 0:31u2 Sc=660ð Þ−1=2 ð4Þ

where u is the wind speed at 10 m above the sea surface and the ratio(Sc/660)−1/2 is used as normalization, Sc is the Schmidt number, theratio of the kinematic viscosity of water to the diffusion coefficient ofCO2, and 660 is the value to which the Schmidt number is normalized

e

f

g

h

6 7 8 9 10

dots), the direction is where the wind and current comes from. b)Wind speedmeasureduth of Gotland. Positive values means sea level is higher north of Östergarnsholm.ensity at 22 m (grey). g) Temperature at 5 m depth (grey) and air temperature (black).

Fig. 4. Total alkalinity, AT (filled triangles) and dissolved inorganic carbon, CT (opencircles) at 5 m depth measured from R/V Skagerak.

5K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

(representing CO2 in seawater at 20 °C). The value of the exponent is aresult from the experimental work by Jähne et al. (1987) andrepresents a wavy, rough surface.

Wanninkhof and McGillis (1999) presented a parameterization ofk with a cubic dependency on wind speed:

kW&McG99 = 0:0283u3 Sc=660ð Þ−1=2 ð5Þ

They argued that this cubic dependency better explains theenhancement of breaking waves and bubbles at high wind speedand also the resistance at low wind speeds by surfactants.

Weiss et al. (2007) presented semi-continuous data of atmospher-ic and oceanic CO2 measured from a platform installation in thesouthern Baltic Sea during about one year. Their data set consisted of7820 half hour measurements for wind speeds up to 20 ms−1. Acombined quadratic and direct wind speed dependency provided thebest fit to their data:

kWeiss07 = 0:365u2 + 0:46u� �

Sc=660ð Þ−1=2 ð6Þ

The choice of wind speed as the controlling variable is reasonabledue to the fact that this parameter is involved inmost processeswhichare likely to influence the transfer, e.g. breaking waves, bubbles andturbulence. However, choosing only one governing variable leads touncertainties in every parameterization.

2.5. Diurnal rates of changes in O2 and CT

To separate changes in O2 and CT related to biological from physicalprocesses (such asmixing and air–sea gas exchange), we calculated thediurnal rates of changes inO2 (dO2/dt) and CT (dCT/dt) in a closed versusan open system. In the closed system, neither advection nor air–sea gasexchange took place, only biological processes occurred. For the opensystem, all three processes took place. To investigate the diurnal patternwe divided the daily rates into a day time part and a night time part.

For the closed system, we used data from the ProductivityAutosampler (PA) and the bottle incubations (BI). From the PA-datawe calculated dO2/dt and from the BI-data we calculated dO2/dt anddCT/dt. The open systemwas represented by data from the moored O2

a

b

Fig. 3. a) Partial pressure of CO2 in the air (pCO2a). b) Partial pressure of CO2 at 5 m depth

(pCO2w) measured with the SAMI buoy (black line) and determined from ship

measurements of AT and CT (dots). Grey line is dissolved oxygen at 5 m, note that they axis is reversed. Lower black line is PAR. The vertical grid lines alternately representday time and night time.

and SAMI sensors. From the O2 sensor we could calculate dO2/dt.Because the SAMI sensor generated pCO2

w we used the programCO2SYS (Lewis and Wallace, 1998) with these pCO2

w data and aconstant AT (1640 μmol L−1) to calculate CT. Eventual uncertainties inthis calculation are of minor importance since we were interested inthe relative changes for CT and not the absolute value.

3. Results

3.1. Physical conditions

Measurements were made in April 2006, and weather conditionswere changing between snow and rain. Sea conditions were roughwith strong currents. A camera investigation of the seafloor at 22 mshowed no soft sediments but a hard stony bottom which alsoindicated the dynamics of the site. The time evolution of themeasurements of physical parameters in Fig. 2 all illustratedoscillations, on hourly as well as on daily time scales. The dominantwind direction was southward, minimizing the land influence for themeasuring site (Fig. 2a). The average wind speed was 6.8 ms−1 andthere were several periods with wind speeds exceeding 10 ms−1

(Fig. 2b). On noon April 6, the wind turned from north–west to southand the wind speed also increased significantly. Fig. 2c illustrates thesea level difference between Landsort and Kungsholmsfort, locationswhich are north and south of Östergarnsholm, respectively. There areno significant tides in the Baltic Sea and these sea level variationswereinstead depending on regional and local winds as well as on largerscale weather system variations over the North Sea and the Balticregion. When the wind speed increased on April 6 the sea leveldifference also increased. The situation was reversed during

Fig. 5. Oxygen saturation state at 5 m depth.

a

b

Fig. 6. a)Difference inpartial pressureofCO2betweenwater andair,ΔpCO2(black) andwindspeed(grey). b)Air seaexchangeofCO2according toWanninkhof (1992) (black),Weiss et al.(2007) (dashed), and Wanninkhof and McGillis (1999) (grey). Negative air sea exchangerepresents CO2 uptake by the sea and positive represents release of CO2 to the atmosphere.

a

b

Fig. 7. Diurnal rates of changes in a) O2 and b) CT. Dashed lines are changes during day time a(BI), triangles are data from the Productivity Autosampler (PA), and circles are data from theand oxygen optodes are open systems.

6 K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

decreasing wind speed as on April 8 and 9. The horizontal current at5 m depth was on average 11 cm s−1 and speed as well as direction(Fig. 2a and d) appeared to be strongly influenced by wind speed,wind direction, and sea level gradients. The low current speed on April6 and 8 was induced by major changes in the wind direction. Changesin the significant wave height were also following wind events closely(Fig. 2e). From continuous measurements of salinity and temperatureat the surface and at the bottom, the density was derived (Fig. 2f) andthe entire water columnwas vertically mixed a couple of times duringthe week. The air temperature varied between 0 and +5 °C and thesea surface temperature was relatively constant, 1.6±0.2 °C (Fig. 2g).This implied an unstable atmospheric stratification during the firstday and stable conditions during the rest of the measuring period.From the continuous PAR measurements (Fig. 2h) we determined theday to begin at 7 a.m. and end at 7 p.m. This was used for thedetermination of the diurnal rates of changes in the in situ O2 and CTmeasurements.

3.2. The partial pressure of CO2 and oxygen

The surface water was constantly under saturated in pCO2 and inthe atmosphere pCO2 was steady around 390 μatm (Fig. 3). The SAMIinstrument sampled pCO2

w data every hour for four days and pCO2w

varied between 246 and 328 μatm with a mean of 291 μatm. From themeasured AT and CT on R/V Skagerak (Fig. 4) the calculated pCO2

w

values ranged between 244 and 304 μatm and were thus generallylower than SAMI data (Fig. 3b). Since the ship measurements weremade at a distance of about 200 m from the SAMI buoy, somedisagreements could be expected. The surface water was supersatu-rated in dissolved O2 and on average the degree of saturation was103% (Fig. 5) and the oxygen concentration varied from 420 to446 μmol L−1 (Fig. 3b). As expected, the variations of O2 and pCO2

w

nd full lines are changes during night time. Squares are data from the bottle incubationsO2 optodes and the SAMI buoy. PA and BI are referred to as closed systems whilst SAMI

a

b

c

Fig. 8. a) Dissolved inorganic carbon, CT (black) and O2 (grey). b) Low pass filter appliedto the anomalies of CT (black) and O2 (grey). c) High pass filter applied to the anomaliesof CT (black) and O2 (grey). Lower grey line in b and c is PAR (unit is μE m−2 s−1). Thevertical grid lines alternately represent day time and night time. Note that the y axis forO2 is reversed.

Fig. 9. Concentrations of phosphate (squares) and nitrate (open circles).

7K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

followed each other closely and were significantly anti-correlatedwith the correlation coefficient r=−0.8.

3.3. Air–sea exchange of CO2

The air–sea exchange of CO2 was calculated using the parameter-izations described in Section 2.4. As expected from the negative pCO2

gradient (Fig. 6a), the flux was constantly directed downwardsinto the sea (Fig. 6b). The parameterized flux had a mean of −0.8±0.7 mmol m−2 h−1 for the kW&McG99 formulation, −1.0±0.6 mmolm−2 h−1 for the kW92 formulation, and −1.3±0.8 mmol m−2 h−1

for the kWeiss07 formulation. Hence, the parameterization which gavethe lowest flux, kW&McG99, was on average 64% lower than theparameterization that gave the largest flux, kWeiss07.

3.4. Daily production and consumption rates of O2 and CT

In the closed system, represented of bottle incubations and the PAsystem, only biological processes were detected. The result showed anO2 production and CT consumption during day time, and an O2

consumption and CT production during night time (Fig. 7). In the opensystem, which were represented of the SAMI and oxygen sensor, bothbiological as well as physical processes contributed to the result. Theresult was similar to the result from the closed system with the mainexception on April 7 during day time when O2 decreased and CTincreased unlike the rest of the days.

4. Discussion

4.1. Dominating processes

The pCO2 under saturated surface water indicated that the springbloom of plankton had started. These measurements are consistentwith previous studies on long time series of pCO2

w in the central BalticSea which also shows an under saturation and decrease in pCO2

w inApril (e.g. Wesslander et al., 2010).

It was in the present study possible to distinguish periods whichwere dominated by air–sea gas exchange, mixing/advection, orbiological processes. All of these processes are present simultaneouslybut in most cases one or two are dominating with less contributionfrom the other. During air–sea gas exchange, the change in pCO2

w isdelayed compared to the change in O2 while there is an immediateresponse during biological activity and mixing/advection, this can beused to distinguish between the processes. For typical oceanconditions, it takes about 20 times longer for a reactive gas, as CO2,to equilibrate compare to a nonreactive gas, as O2 (e.g. Emerson andHedges, 2009). This relation is however not applicable to the BalticSea because of the brackish condition. The equilibration time (τ) is thetime it takes for amixed water layer to be equilibrated after an air–seagas exchange. For O2, τ can be estimated as τO2=MLD/k, where MLDis themixed layer depth and k is themean gas exchangemass transfercoefficient (5 md−1). For a 5 m thick layer, O2 hence take 1 day toequilibrate. For CO2, τ is estimated as τCO2=MLD/k×CT/ [CO2]×1/RF,where [CO2] is the CO2 concentration and RF is the Revelle factor.When we use values from this investigation, τCO2=3.2 days for a 5 mthick layer, whichmeans that CO2 takes 3.2 times longer to equilibratecompare to O2 (valid for Baltic Sea conditions in April 2006). Themainreason for this faster relative response is that the RF in the Baltic Sea ishigher than in the ocean because AT~CT.

To emphasize periods dominated by different processes CT and O2

were filtered (Fig. 8a–c). For this analysis CT was used instead of pCO2w

since CT is not sensitive to temperature. We used pCO2w from the SAMI

buoy and a constant AT to calculate CT, see also Section 2.5. A low passfilter which removes the daily signal (Fig. 8b) and a high pass filterwhich keeps daily and shorter cycles (Fig. 8c) was applied. This

method has previously been successful in analyzing diel cycles of e.g.nitrate (Johnson et al., 2006).

In the beginning of our measurements (April 5–6) mixing seemedto play a dominant role with no clear trends between day and nightand an immediate coupling between changes in pCO2

w and O2. Thiswas particularly evident on the night between April 5 and 6 whenpCO2

w decreased as O2 increased. According to the low pass filtereddata, this first period was controlled by cycles much longer than thedaily time scale which supports the idea that mixing prevailed(Fig. 8b). On April 7, from about 2 a.m. to 9 p.m., pCO2

w increased andO2 decreased, contrary to what would be the case if the variation wascontrolled of only biological production. There was also a delay of

Table 1The mole O2:CT ratio during day time versus night time. Ratios were determined withO2 and CT measurements from the bottle incubations.

Date O2:CT day O2:CT night

April 6, 2006 −1.11 −1.17April 7, 2006 −1.11 −1.0April 8, 2006 −0.88 −1.18April 9, 2006 −0.96 −0.91

8 K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

about 2 h between the O2 decrease and pCO2w increase which indicates

that air–sea gas exchange took place and was the dominating factor.During this event, CO2 was hence taken up by the sea and O2 wasdegassed. A clear CO2 influx was also calculated during this period(Fig. 6b), the period was also characterized of higher wind speed andmore wave activity. On the night between April 7 and 8, mixing/advection were probably again the dominating factor since O2 were ata constant level and the pCO2

w were oscillating. The high pass filtereddata (Fig. 8c) show daily and shorter cycles during the whole periodbut it is only in the end (April 8–10) that the variations seem to have aclear biological connection with decreasing pCO2

w (CT) and increasingO2 during daylight and vice versa during the night. This, together withthe immediate response between pCO2

w and O2, indicates thatbiological processes were dominant for this period.

The temporal evolution of AT and nutrients also illustrated how thevariabilitywasdeterminedbydifferentprocesses (Figs. 4 and9). BetweenApril 6 and 8 there was a major decrease in both AT and nutrients whichthen again increased. The timing for these particular changes agree wellwith the period considered to be dominated by mixing and air–seaexchange. Further, it was observed that both phosphate and nitrateconcentrationswere reduced compared towinter values 2006 (accordingto the monitoring program by SMHI). This also substantiate that thenutrient consuming primary production had started.

It was clear from the bottle incubations and the PA system thatphotosynthesis occurred during day time and respiration occurredduring night time for the whole period. When looking at the opensystem, however, mixing and air–sea exchange contributed to the O2

and CT variations. This was particularly evident on April 7 during daytime when O2 decreased and CT increased unlike the rest of the days(Fig. 7).

In the bottle incubations, both O2 and CT were determinedallowing the O2:CT ratio determination (Table 1). Overall, the O2:CTratio was less than the Redfield ratio, during both day and night. Theaverage ratio in this study was −1.0 compared to the Redfield O2:CTratio that is −1.3 (Redfield et al., 1963).

4.2. Air–sea gas exchange

Ifwe take themeanparameterizedflux for thewholeperiodusing thekW92 formulation, −1.0±0.6 mmol m−2 h−1, and allow this to be validfor whole of April the flux would be −691±432 mmol m−2 month−1.According to data used in Wesslander et al. (2010) the monthly flux atthe Gotland Sea (BY15) in April 2006 was −386 mmol m−2 month−1.The latter flux estimate was also based on the kW92 formulation butcalculated from3-hwind speed togetherwithmonthlymeasurements ofpH and AT from the SMHImonitoring program. These two flux estimatesillustrates the variations caused by the different methods used as well asthe spatial variability of the CO2 flux. One conclusion from thiscomparison is the need for high resolution measurements of pCO2

w inboth time and space.

5. Summary and conclusions

Several chemical and physical parameters, in both the air and inthe sea, were, for one week in April 2006, measured simultaneously

using a unique multitechnology approach to provide an overview ofthe CO2/O2 system in coastal Baltic Sea surface water.

Sea surface pCO2w and O2 were significantly anti-correlated and

highly variable. These temporal variations in pCO2w and O2 illustrated

the patchiness of biological production, how the processes vary inboth time and space, as well as how they vary with mixing andadvection. We were able to distinguish periods which weredominated with air–sea gas exchange, mixing/advection, as well asof biological processes.

The high variability was also observable in the parameterized air–sea gas exchange of CO2. Since the surface water was constantly undersaturated with respect to CO2, there was an inflow of CO2 to the sea.The average CO2 uptake was −0.8 mmol m−2 h−1 with the kW&McG99

formulation, −1.0 mmol m−2 h−1 with the kW92 formulation, and−1.3 mmol m−2 h−1 with the kWeiss07 formulation.

High resolution data give important information about processesand variabilities. A question arisen from this study is how one shouldinterpret e.g. marine monitoring measurements, which often are notavailable more than once a month. Estimates of e.g. CO2 air–seaexchange, trends in pH (ocean acidification) and nutrient content areoften based on these monthly measurements. We would thereforelike to recommend more studies with high frequency measurements,e.g. from a permanent marine observatory. To resolve the involvedprocesses one would probably need to measure every hour.

Acknowledgements

This work is a part of the GEWEX/BALTEX and the BONUS/Baltic-Cprogrammes and has been funded by the University of Gothenburgand its Marine Research Centre, the Swedish Research Council (VR)under the G 600-335/2001 contract and the Swedish Research Councilfor Environment, Agricultural Sciences and Spatial Planning (FOR-MAS) under contract 21.0/2004-0374. Björn Carlsson and HansBergström (Uppsala University) are acknowledged for the practicalhelp with the meteorological measurements and data processing. Wewould also like to give our appreciation to Sara Jutterström, DanielaHanslik, Elin Almroth, Leif G. Anderson, and Ann-Sofie Smedman fortheir contributions, and to the crew of R/V Skagerak.

References

Borges, A.V., Frankignoulle, M., 1999. Daily and seasonal variations of the partialpressure of CO2 in surface seawater along Belgian and southern Dutch coastal areas.J. Mar. Syst. 19, 251–266.

Borges, A.V., Schiettecatte, L.-S., Abril, G., Delille, B., Gazeau, F., 2006. Carbon dioxide inEuropean coastal waters. Estuar. Coast. Shelf Sci. 70, 375–387.

Bozec, Y., Thomas, H., Elkalay, K., de Baar, H.J.W., 2005. The continental shelf pump forCO2 in the North Sea—evidence from summer observation. Mar. Chem. 93,131–147.

Bozec, Y., Thomas, H., Schiettecatte, L.-S., Borges, A.V., Elkalay, K., de Baar, H.J.W., 2006.Assessment of the processes controlling seasonal variations of dissolved inorganiccarbon in the North Sea. Limnol. Oceanogr. 51 (6), 2746–2762.

Brunnegård, J., Grandel, S., Ståhl, H., Tengberg, A., Hall, P., 2004. Nitrogen cycling indeep-sea sediments of the Porcupine Abyssal Plain, NE Atlantic. Prog. Oceanogr. 63,159–181.

Carlberg, S.R., 1972. New Baltic Manual, With Methods for Sampling and Analysis ofPhysical, Chemical and Biological Parameters, International Council for theexploration of the Sea, Series A, No 29.

DeGrandpre, M.D., Hammar, T.R., Smith, S.P., Sayles, F.L., 1995. In situ measurements ofseawater pCO2. Limnol. Oceanogr. 40, 969–975.

DeGrandpre, M.D., Baehr, M.M., Hammar, T.R., 1999. Calibration-free optical chemicalsensors. Anal. Chem. 71, 1152–1159.

DeGrandpre, M.D., Baehr, M.M., Hammar, T.R., 2000. Development of an opticalchemical sensor for oceanographic applications: the Submersible AutonomousMoored Instrument for Seawater CO2. In: Varney, M. (Ed.), Chemical Sensors inOceanography. Gordon and Breach publ., Amsterdam, pp. 123–141.

Emerson, S.R., Hedges, J.I., 2009. Chemical Oceanography and the Marine Carbon Cycle.Cambridge University Press.

Gattuso, J.-P., Frankingoulle, M., Wollast, R., 1998. Carbon and carbonate metabolism incoastal aquatic ecosystems. Annu. Rev. Ecol. Syst. 29, 405–434.

Haraldsson, C., Anderson, L.G., Hassellöv, M., Hulth, S., Olsson, K., 1997. Rapid, high-precision potentiometric titration of alkalinity in the ocean and sediment porewaters. Deep Sea Res. I 44, 2031–2044.

9K. Wesslander et al. / Journal of Marine Systems 86 (2011) 1–9

Högström, U., Sahlée, E., Drennan, W., Kahma, K.K., Smedman, A., Johansson, C.,Pettersson, H., Rutgersson, A., Tuomi, L., Zhang, F., Johansson, M., 2008. Momentumfluxes and wind gradients in the marine boundary layer—a multi platform study.Boreal Environ. Res. 13, 475–502.

Jähne, B., Münnich, K.-O., Bösinger, R., Dutzi, A., Huber, W., Libner, P., 1987. On theparameters influencing air–water gas exchange. J. Geophys. Res. 92, 1937–1949.

Johnson, K.M., Sieburth, J.M., Williams, P.J., Brändström, L., 1987. Coulometric totalcarbon dioxide analysis for marine studies: automation and calibration. Mar. Chem.21, 117–133.

Johnson, K.S., Coletti, L.J., Chavez, F.P., 2006. Diel nitrate cycles observed with in situsensors predict monthly and annual new production. Deep Sea Res. I 53, 561–573.

Kuss, J., Roeder, W., Wlost, K.P., 2006. Time-series of surface water CO2 and oxygenmeasurements on a platform in the central Arkona Sea (Baltic Sea): seasonality ofuptake and release. Mar. Chem. 101, 220–232.

Langdon, C., 1984. Dissolved oxygen monitoring system using a pulsed electrode:design, performance, and evolution. Deep Sea Res. A 31 (11), 1357–1367.

Langdon, C., Marra, J., Knudson, C., 1995. Measurements of net and gross O2 production,dark O2 respiration, and 14C assimilation at the marine light-mixed layers site(59°N, 21°W) in the northeast Atlantic Ocean. J. Geophys. Res. Oceans 100 (4),6645–6653.

Lewis, E., Wallace, D.W.R., 1998. Program Developed for CO2 Systems Calculations.ORNL/CDIAC-105, Carbon Dioxide Information Analysis Centre, Oak Ridge NationalLaboratory, U.S. Department of Energy, Oak Ridge, Tennessee.

Liss, P.S., Chuck, A.L., Turner, S.M., Watson, A.J., 2004. Air–sea gas exchange in Antarcticwaters. Antarct. Sci. 16, 517–529.

Murphy, I., Riley, J.P., 1962. A modified single solution method for the determination ofphosphate in natural waters. Anal. Chim. Acta 27, 31–36.

Nordforsk Miljövårdssekretariatet, 1973. Publikation 1973:3, Interkalibrering avmetoder för bestämning av nitrat och totalnitrogen.

Omstedt, A., Gustafsson, E., Wesslander, K., 2009. Modelling the uptake and release ofcarbon dioxide in the Baltic Sea surface water. Cont. Shelf Res. 29, 870–885.

Redfield, A.C., Ketchum, B.H., Richards, F.A., 1963. The influence of organisms on thecomposition of sea-water. In: Hill, M.N. (Ed.), The composition of seawater.Comparative and descriptive oceanography: The sea: ideas and observations onprogress in the study of the seas, 2, pp. 26–77.

Rutgersson, A., Smedman, A., Omstedt, A., 2001. Measured and simulated latent andsensible heat fluxes at two marine sites in the Baltic Sea. Bound. Lay. Meteorol. 99,53–84.

Rutgersson, A., Norman, M., Schneider, B., Pettersson, H., Sahlée, E., 2008. The annualcycle of carbon dioxide and parameters influencing the air–sea carbon exchange inthe Baltic Proper. J. Mar. Syst. 74, 381–394.

Schneider, B., Nausch, G., Nagel, K., Wasmund, N., 2003. The surface water CO2 budget forthe Baltic Proper: a new way to determine nitrogen fixation. J. Mar. Syst. 42, 53–64.

Schneider, B., Kaitala, S., Maunula, P., 2006. Identification and quantification of planktonbloom events in the Baltic Sea by continuous pCO2 and chlorophyll a measure-ments on a cargo ship. J. Mar. Syst. 59, 238–248.

Schneider, B., Sadkowiak, B., Wachholz, F., 2007. A new method for continuousmeasurements of O2 in surface water in combination with pCO2 measurements:implications for gas phase equilibration. Mar. Chem. 103 (1–2), 163–171.

Ståhl, H., Hall, P., Tenberg, A., Josefson, A., Streftaris, N., Zenetos, A., Karageorgis, A.P.,2004. Respiration and sequestering of organic carbon in shelf sediments of theoligotrophic northern Aegean Sea. Mar. Ecol. Prog. Ser. 269, 33–48.

Strickland, J.D.H., Parsons, T.R., 1968. A practical handbook of sea water analysis.Fisheries Research Board of Canada: Bulletin, 167.

Svensk Standard, SS028133, 1991. Vattenundersökningar – bestämning av summan avnitrit-och nitrat-nitrogen i vatten.

Tengberg, A., Almroth, E., Hall, P., 2003. Resuspension and its effects on organic carbonrecycling and nutrient exchange in coastal sediments: in situ measurements usingnew experimental technology. J. Exp. Mar. Biol. Ecol. 285–286, 119–142.

Tengberg, A., Hovdenes, J., Andersson, H.J., Brocandel, O., Diaz, R., Herbert, D., Arnerich,T., Huber, C., Körtzinger, A., Khripounoff, A., Rey, F., Rönning, C., Schimanski, J.,Sommer, S., Stangelmayer, A., 2006. Evaluation of a lifetime-based optode tomeasure oxygen in aquatic systems. Limnol. Oceanogr. Meth. 4, 7–17.

Thomas, H., Schneider, B., 1999. The seasonal cycle of carbon dioxide in Baltic Seasurface waters. J. Mar. Syst. 22, 53–67.

Wanninkhof, R., 1992. Relationship between wind speed and gas exchange over theocean. J. Geophys. Res. 97 (C5), 7373–7382.

Wanninkhof, R., McGillis, W.R., 1999. A cubic relationship between air–sea CO2

exchange and wind speed. Geophys. Res. Lett. 26 (13), 1889–1892.Weiss, R.F., 1974. Carbon dioxide in water and seawater: the solubility of a non-ideal

gas. Mar. Chem. 2, 203–215.Weiss, A., Kuss, J., Peters, G., Schneider, B., 2007. Evaluating transfer velocity-wind

speed relationship using a long-term series of direct eddy correlation CO2 fluxmeasurements. J. Mar. Syst. 66, 130–139.

Wesslander, K., Omstedt, A., Schneider, B., 2010. Inter-annual and seasonal variations inthe air–sea CO2 balance in the central Baltic Sea and the Kattegat. Cont. Shelf Res.30, 1511–1521.

Williams, P.J.L.E.B., Jenkinson, N.W., 1982. A transportable microprocessor-controlledprecise Winkler titration suitable for field station and shipboard use. Limnol.Oceanogr. 27 (3), 576–584.

Winsor, P., Rodhe, J., Omstedt, A., 2001. Baltic Sea ocean climate: an analysis of 100 yr ofhydrographic data with focus on the freshwater budget. Clim. Res. 18, 5–15.

Winsor, P., Rodhe, J., Omstedt, A., 2003. Erratum: Baltic Sea ocean climate: an analysis of100 yr of hydrographic data with focus on the freshwater budget. Clim. Res. 18,5–15 Clim. Res., 25, 183.

Wootton, J.T., Pfister, C.A., Forester, 2008. Dynamic patterns and ecological impacts ofdeclining ocean pH in a high-resolution multi-year dataset. PNAS 105 (48),18848–18853.

Web references

http://www.aadi.nohttp://www.smhi.sehttp://www.sunburstsensors.com