Inter-annual variation of the air-sea CO2 balance in the southern Baltic Sea and the Kattegat

Page 1

Continental Shelf Research 30 (2010) 1511–1521

Contents lists available at ScienceDirect

Continental Shelf Research

0278-43

doi:10.1

n Corr

E-m

journal homepage: www.elsevier.com/locate/csr

Inter-annual and seasonal variations in the air–sea CO2 balance in the centralBaltic Sea and the Kattegat

Karin Wesslander a,n, Anders Omstedt a, Bernd Schneider b

a University of Gothenburg, Department of Earth Sciences, Box 460, SE-405 30 Goteborg, Swedenb Baltic Sea Research Institute, Department of Marine Chemistry, Seestrasse 15, D-18119 Rostock-Warnemunde, Germany

a r t i c l e i n f o

Article history:

Received 19 October 2009

Received in revised form

25 May 2010

Accepted 27 May 2010Available online 4 June 2010

Keywords:

Carbon cycle

Carbon dioxide

Air–sea exchange

Baltic Sea

Kattegat

43/$ - see front matter & 2010 Elsevier Ltd. A

016/j.csr.2010.05.014

esponding author. Tel.: +46 0 317862873.

ail address: [email protected] (K. W

a b s t r a c t

We estimated the net annual air–sea exchange of carbon dioxide (CO2) using monitoring data from the

East Gotland Sea, Bornholm Sea, and Kattegat for the 1993–2009 period. Wind speed and the sea surface

partial pressure of CO2 (pCO2w), calculated from pH, total alkalinity, temperature, and salinity, were used

for the flux calculations. We demonstrate that regions in the central Baltic Sea and the Kattegat alternate

between being sinks (�) and sources (+) of CO2 within the �4.2 to +5.2 mol m�2 yr�1 range. On

average, for the 1994–2008 period, the East Gotland Sea was a source of CO2 (1.64 mol m�2 yr�1), the

Bornholm Sea was a source (2.34 mol m�2 yr�1), and the Kattegat was a sink (�1.16 mol m�2 yr�1).

Large inter-annual and regional variations in the air–sea balance were observed. We used two

parameterizations for the gas transfer velocity (k) and the choice varied the air–sea exchange by a factor

of two. Inter-annual variations in pCO2w between summers were controlled by the maximum

concentration of phosphate in winter. Inter-annual variations in the CO2 flux and gas transfer velocity

were larger between winters than between summers. This indicates that the inter-annual variability in

the total flux was controlled by winter conditions. The large differences between the central Baltic Sea

and Kattegat were considered to depend partly on the differences in the mixed layer depth.

& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The total CO2 inventory of the oceans exceeds that of theatmosphere by a factor of approximately 50 (Solomon et al., 2007).This implies that small changes in the ocean CO2 might havedrastic consequences for the atmospheric CO2 budget. AtmosphericCO2 concentrations stabilized at approximately 280 matm after thelast glaciation period, indicating balance in the global carbon cycleand between oceanic sinks and sources. For approximately 200years, anthropogenic CO2 emissions have been superimposed onthe natural cycling of CO2 between the oceans and the atmosphere,perturbing the pre-industrial steady state. This perturbation hasresulted in a net oceanic CO2 uptake, thus dampening theanthropogenic CO2 increase in the atmosphere but also reducingocean pH, which may be referred to as ocean acidification (e.g.,Caldeira and Wickett, 2003; Sabine et al., 2004).

The cold, productive waters at high latitudes are the mainsinks of atmospheric CO2, whereas upwelling regions constitutethe major sources (Takahashi et al., 1993). The importance ofcoastal seas, for both the natural CO2 cycle and the uptake/releaseof anthropogenic CO2, has been paid more attention to in recentyears. Although shelf regions and marginal seas account for a

ll rights reserved.

esslander).

small part of the ocean surface, their share of global marineprimary production is high. Since biological activity constitutes amajor factor controlling the state of the marine CO2 system (e.g.,in the Baltic Sea; Thomas and Schneider, 1999), these regions playan important role in CO2 cycling. Whether the sea acts as a sink orsource of CO2 depends greatly on the relationship between the netproduction and mineralization of organic carbon. When miner-alization exceeds biological production, for example, because of alarge input of organic matter from land, CO2 is released into theatmosphere. In contrast, CO2 is taken up when production exceedsmineralization. Shelf waters that have taken up CO2 from theatmosphere may be transported into deeper ocean water layers.This process, which is driven by enhanced cooling of theshallower shelf zone, is called the ‘continental shelf pump’(Tsunogai et al., 1999).

Some studies have examined the CO2 balance of the Europeanshelf regions (e.g., Borges et al., 2006; Chen and Borges, 2009),finding that the continental shelves are sinks while the analyzedestuaries act as sources of atmospheric CO2. A recent model study(Omstedt et al., 2009) indicates that the Baltic Sea served as asource of atmospheric CO2 during pre-industrial times. The studyalso indicates that during industrialization and the subsequentincreasing eutrophication, seasonal variability increased, makingthe Baltic Sea both a sink and source of atmospheric CO2.According to Thomas et al. (2004), the North Sea acts as a netsink of 1.4 mol m�2 yr�1, and Thomas and Schneider (1999) state

Page 2

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–15211512

that the central Baltic Sea acts as a net sink of 0.9 mol m�2 yr�1.On the other hand, the Gulf of Bothnia acts as a source of3.1 mol m�2 yr�1 (Algesten et al., 2004). Uncertainties in theseestimates are associated both with insufficient spatial andseasonal resolution of the CO2 measurements and with question-able parameterizations of the gas exchange transfer velocity (e.g.,Rutgersson et al., 2008; Wanninkhof, 1992).

Our study re-evaluates the role of the Baltic Sea as a sink orsource of atmospheric CO2 and, in particular, assesses theseasonal and inter-annual variability of the air–sea CO2 balance.We have used long-term monitoring data for pH, total alkalinity(AT), temperature, and salinity to calculate the surface water CO2

partial pressure, pCO2w, which is then used to estimate the CO2

flux between air and sea in the central Baltic Sea and the Kattegat.We are aware that calculating the pCO2

w based on AT and pHmeasurements is difficult and may cause biases. Still, we thinkthat, due to the exceptional length of the time series, the datacontain valuable information about the Baltic Sea CO2 system.

This study also contributes to the understanding and devel-opment of models of the Baltic Sea carbon system. The studyexamines several important factors, such as seasonal, regional,and inter-annual differences, all of which must be considered infuture model simulations.

2. Materials and methods

2.1. Study site

The Baltic Sea is a brackish semi-enclosed sea on the continentalshelf with a river water input of about 15,000 m3 s�1 (Bergstrom andCarlsson, 1994) and a net precipitation rate of about 1500 m3 s�1

Fig. 1. The Baltic Sea and the Kattegat. The black line is the route of the cargo ship Finnp

the monitoring stations in the Kattegat, Bornholm Sea, and East Gotland Sea, respectiv

(Omstedt et al., 2004). This fresh water brings large amountsof nutrients and inorganic and organic carbon (Hjalmarssonet al., 2008; Morth et al., 2007; Omstedt et al., 2004). The system isdynamic, resulting in horizontal and vertical gradients in variablesthat control the carbon dioxide system, such as pH, AT, salinity, andtemperature (for areas in the Baltic Sea with different AT values, seeHjalmarsson et al., 2008). In analyzing the CO2 balance, we focused onthree regions: the East Gotland Sea and Bornholm Sea, both of whichare located in the central Baltic Sea, and the Kattegat. In each of thesethree regions there is a monitoring station where pH, AT, temperature,and salinity are regularly measured at monthly intervals. The EastGotland Sea and the Bornholm Sea were represented by themonitoring stations BY15 and BY5, respectively, and Kattegat by theAnholt East (AnE) station (Fig. 1). BY15 and BY5 are at the centralpoints in the East Gotland and Bornholm Sea basins, respectively,while AnE in the Kattegat is located in the Baltic Sea sill region. Thehydrography of the Baltic Sea has been described, for example, inFonselius and Valderrama (2003).

2.2. The atmospheric data

To obtain a continuous function for the partial pressure of CO2

in the atmosphere (pCO2a), we fitted a cosine function to the

available data for the mole fraction of CO2 in dry air, xCO2. The

available data we used for this comprised xCO2measurements of

from the Polish coast, obtainable from the World Data Centre forGreenhouse Gases. In these data and for the studied period, aseasonal variation of 710 ppm and an annual increase of1.9 ppm yr�1 were observed. Using this information, we fittedthe following function (Eq. (1)):

xCO2¼ exþacos½bðx�cÞ�þd ð1Þ

artner between Lubeck in Germany and Helsinki in Finland. AnE, BY5, and BY15 are

ely.

Page 3

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–1521 1513

where e is the annual increase of 1.9 ppm yr�1, a the seasonalamplitude of 10 ppm, b the period (2p/12), c¼2 the phase shift(xa

CO2has its maximum in February), and d is the start value (d is

358 ppm when starting the time series in January 1993). Becausethis calculated time series is based on the mole fraction of CO2, weaccounted for the water vapour pressure at the sea surface andthe atmospheric pressure in order to obtain pCO2

a (Weiss andPrice, 1980).

To determine the gas exchange transfer velocity, we used thegeostrophic wind speed from the SMHI 11�11 gridded databasefor each of the Gotland Sea, Bornholm Sea, and Kattegat (availablefrom the BALTEX Hydrological Data Centre). This dataset has beencompared with data from the ERA40 reanalysis project and foundto agree with it well (Omstedt et al., 2005). The geostrophic windspeed was reduced to represent the wind speed at 10 m, usingreduction formulas according to Omstedt and Axell (2003). Thereduction formulas were developed based on wind speedmeasurements made at Ostergarnsholm, a small island near theeastern coast of Gotland where Uppsala University has ameteorological station (see Hogstrom et al., 2008, for a descrip-tion of the station). We compared observed wind speed fromOstergarnsholm with the calculated wind speed from SMHI11�11 (Fig. 2). The comparison indicated generally goodagreement, though the Ostergarnsholm winds are lower at highwind speeds, probably due to land friction influence. We used thedaily average wind speed for the calculations. The wind field inthe Baltic Sea is highly variable, and the various air–sea fluxes(i.e., momentum, heat, and gas) are often non-linearly related tothe wind speed. It is therefore important to use at least daily windspeeds to resolve the fluxes in the Baltic Sea.

2.3. Monitoring data

For this study, we used measurements of pH, AT, salinity, andtemperature for the 1993–2009 period. These measurements arepart of the Swedish marine monitoring program and areobtainable from the Swedish oceanographic data centre, which

Fig. 2. Observations of daily mean wind speed at Ostergarnsholm versus

calculated daily wind speed from the SMHI 11�11 gridded database valid for

the East Gotland Basin.

is maintained by the Swedish Meteorological and HydrologicalInstitute (SMHI). The carbon system can be explained by any twoof the four parameters pH, AT, pCO2

w, and CT (total CO2). Of thesefour parameters, only pH and AT are measured continuously at aseasonal resolution in the Baltic Sea and Kattegat. SMHI, whichhas made the measurements, is accredited by the Swedish Boardfor Accreditation and Conformity Assessment (SWEDAC) accord-ing to SS-EN ISO/IEC 17025. Data resolution is monthly at BY15and BY5 and bimonthly at AnE. All data used are from 5 m belowthe sea surface. Salinity and temperature were measured with aCTD probe. The pH was determined using NBS buffer (Grasshoffet al., 1999; ISO 10523, 1994) and the total alkalinity wasdetermined using potentiometric titration (Grasshoff et al., 1999).We also used SMHI measurements of inorganic phosphate thatwere made using flow analysis and spectrophotometric detection(Grasshoff et al., 1999). The uncertainties in the pH and AT

measurements were 70.03 pH units and 75%, respectively. Theuncertainties for phosphate were 712% in the 0.02–0.2 mMconcentration interval and 74% in the 0.2–10 mM concentrationinterval. Notable, these reported uncertainties are expandeduncertainties with a standard coverage factor of two, which inpractice, means that the uncertainties can be divided by two.Observe that these reported uncertainties are not equivalent tothe repeatability, which is considered even better.

2.3.1. Temperature and salinity

The differences between the sea surface temperatures (SSTs) inthe three regions were small and confined to the summer peaks(Fig. 3a). However, large inter-annual variations were observed,particularly for the summer maximum SSTs, which were stronglycorrelated between the three regions. The surface water salinitydistribution is controlled by the strong river water input in thenorth/east of the Baltic Sea, resulting in a strong increase in trendtowards the Kattegat (Fig. 3b and c). The seasonality of the salinitywas most pronounced at station BY15, being characterized by lowsummer values, due to the development of a shallow thermocline.

2.3.2. pH and total alkalinity

High pH values were observed in spring/summer due to CO2

consumption by intense biological production (Fig. 4a). In winter,pH decreased when CO2 was brought back to the surface waterthrough mineralization and mixing. This seasonal variation waslargest in the East Gotland Sea and smallest in the Kattegat. Inseveral years, the central Baltic Sea had two summer pH maximaseparated by one or two months. These maxima were attributedto two production periods, the spring bloom and the laternitrogen-fixing cyanobacteria bloom (Schneider et al., 2009).The inter-annual variations in pH occurred mainly betweensummers, probably due to variations in biological activity. Therewas one period, 1993–1998, marked by decrease in pH values insummer; if this negative trend depended on less biologicalactivity, it would then indirectly depend on a smaller supply ofavailable nutrients. According to Stigebrandt and Gustafsson(2007), the phosphate concentration in the Baltic Proper surfacelayer was halved in the 1991–1997 period, a phenomenon theybelieve is coupled to the major deep water inflow of 1993–1994.This illustrates how the renewal of Baltic Sea deep water alsoaffects the surface water. The inter-annual pH variations in winterare relatively smaller and most likely due to variations in thevertical mixing that enriches the surface water with CO2.

The total alkalinity data for the three stations are presented inFig. 4b and c. In Kattegat, AT was strongly correlated to salinity(Fig. 5) and was much higher than in the central Baltic Sea becauseof the oceanic influence (higher salinity). In the central Baltic Sea, AT

changes were not that closely coupled to salinity changes (Fig. 5).

Page 4

Fig. 3. (a) Temperature and (b,c) salinity at a depth of 5 m. Data are from the monitoring stations in the Kattegat (Anholt East—black circles) and the central Baltic Sea (BY5

in the Bornholm Sea—grey squares, BY15 in the East Gotland Sea—open circles).

Fig. 4. (a) pH according to the NBS scale and (b,c) total alkalinity, AT, in mmol kg–1, at a depth of 5 m. Data are from the monitoring stations in the Kattegat (Anholt

East—black circles) and the central Baltic Sea (BY5 in the Bornholm Sea—grey squares, BY15 in the East Gotland Sea—open circles).

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–15211514

Page 5

Fig. 5. Total alkalinity, AT, versus salinity for data from a depth of 5 m at the

monitoring stations in the Kattegat (black circles), Bornholm Sea (grey squares),

and East Gotland Sea (open circles).

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–1521 1515

Instead, AT variations were probably associated with the advectionof surface water from areas with different alkalinities stemmingfrom different AT contents in the inflowing rivers (Hjalmarsson et al.,2008). We have studied a period of 17, years which is a short periodfor the identification of trends in view of the high inter-annualvariability. Nevertheless, the AT in both Kattegat and the centralBaltic Sea has increased over the 1993–2009 period. As the sourcesof AT in the Baltic Sea are so different from each other, with highvalues coming from the Gulf of Finland and Riga and low valuesfrom the Gulf of Bothnia, one explanation could be an internalchange in the Baltic Sea circulation. It is also conceivable that thelateral mixing pattern has changed, such that high AT river waterfrom continental Europe has increasingly been mixed into thecentral Baltic Sea, affecting the Kattegat as well.

2.4. Calculation of the partial pressure of CO2

With pH, AT, salinity, and temperature data available, it ispossible to calculate the partial pressure of CO2 in the water (pCO2

w).We did this using the standard CO2SYS program (Lewis and Wallace,1998) with the following equilibrium constants: K0 (Weiss, 1974), K1

and K2 (Mehrbach et al., 1973 as refitted by Dickson and Millero,1987, 1989, for Kattegat and for the central Baltic Sea by Milleroet al., 2006 because of low salinities), KB (Dickson, 1990), and KW

(Millero,1995). The calculated pCO2w values were also compared with

direct pCO2w measurements. The Leibniz Institute for Baltic

Sea Research, Warnemunde, Germany (IOW—Institut furOstseeforschung Warnemunde) has made continuous measure-ments of pCO2

w at a depth of 5 m onboard the cargo vesselFinnpartner. This ship crosses between Lubeck and Helsinki at atwo-day interval, alternately crossing the eastern and westernGotland Sea (Schneider et al., 2006, 2009) (Fig. 1). Data fromFinnpartner were recorded from July 2003 to December 2005.

2.5. The air–sea exchange of CO2

The air–sea exchange of CO2, FCO2, is controlled by the wind

speed-dependent gas transfer velocity, k, the temperature- and

salinity-dependent solubility constant, K0, and the difference inpCO2 between water and air (pCO2

w–pCO2a¼DpCO2) (Eq. (2)):

FCO2¼ kK0ðpCOw

2 �pCOa2Þ ð2Þ

The expression for k as a function of wind speed is notstandardized and several parameterizations exist. To illustrate thedifferent results they give, we chose to calculate the air–sea fluxusing two parameterizations: kW92 from Wanninkhof (1992)(Eq. (3)) and kLM86 from Liss and Merlivat (1986) (Eq. (4)).Wanninkhof (1992) was chosen because it is a commonparameterization that facilitates comparison with other studies.Liss and Merlivat (1986) was added because their parameteriza-tion has been demonstrated to be suitable for the Baltic Sea due tothe lower fetch area and the reduction in k caused by organicfilms on the sea surface (unpublished data by Schneider).

kW92 ¼ 0:31u2

ffiffiffiffiffiffiffiffiffi660

Sc

sð3Þ

KLM86 ¼ 0:17u

ffiffiffiffiffiffiffiffiffi660

Sc

rur3:6

KLM86 ¼ ð2:85u�9:65Þ

ffiffiffiffiffiffiffiffiffi660

Sc

r3:6our13

KLM86 ¼ ð5:9u�49:3Þ

ffiffiffiffiffiffiffiffiffi660

Sc

ru413

ð4Þ

where k is the gas transfer velocity (cm h�1), u the wind speed(m s�1), and Sc the dimensionless Schmidt number for fresh water(Wanninkhof, 1992). As both of these parameterizations are validfor short-term wind speeds, we have used daily mean windspeeds in the calculations. The wind field in the Baltic Sea ishighly variable, so we use these high-time-resolution wind datato increase the quality of our calculations. Accordingly, pH, AT,salinity, and temperature were linearly interpolated to obtaindaily values for K0, Sc, and pCO2

w, and thus for the CO2 fluxes. Withthis method, we are accounting for the co-variation of the meanwind speed and the seasonality of pCO2

w, which may lead to aconsiderable bias in the flux calculations when averaging thevariables in Eq. (2) over longer time intervals. Daily FCO2

valueswere then calculated with Eq. (2) using the corresponding valuesof k, K0, Sc, and pCO2. From the daily flux calculations, the annualnet air–sea exchange of CO2 was calculated for the East GotlandSea, Bornholm Sea, and Kattegat using the two parameterizationsof k. Because too few measurements of pH and AT were availablefor 1993, no CO2 budget was made for that year.

3. Results

3.1. The calculated partial pressure of CO2

3.1.1. Seasonal variations

The pCO2w in the surface water largely reflected the pH, so when

pH dropped the pCO2w increased and vice versa (Figs. 4a and 6). In

the central Baltic Sea, the surface water was supersaturated withCO2 in winter and unsaturated in summer (Fig. 6a and b). In theKattegat, the surface water was more often unsaturated and thewinter pCO2

w was below the maxima in the central Baltic Sea(Fig. 6c). In early spring, pCO2

w dropped and reached unsaturatedlevels that lasted until early autumn. The two summer maxima seenin the pH data were also found in the pCO2

w data in the central BalticSea, but here they appeared as minima.

The differences between the East Gotland Sea and BornholmSea were seen mainly in summer. Summer values were lower inthe Gotland Sea than in the Bornholm Sea, while the winter values

Page 6

Fig. 7. Calculated pCO2w in the East Gotland Sea, filled circles, and measured pCO2

w

from the Finnpartner cargo ship, open circles.

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–15211516

were almost equal. Moreover, distinct pCO2w seasonality was not

detectable in the Kattegat (Fig. 6c).

3.1.2. Inter-annual variations

The central Baltic Sea and Kattegat experienced inter-annualvariations of pCO2

w. In the central Baltic Sea, the pCO2w low in summer

varied between 50 matm (in the East Gotland Sea) and 273 matm (inthe Bornholm Sea), while the pCO2

w high in winter varied between607 matm (in the East Gotland Sea) and 470 matm (in the BornholmSea). An increase in trend for the pCO2

w values was detected at allthree stations over the 1993–1998 period (Fig. 6), which is consistentwith the corresponding decrease in pH. Throughout the 1993–2009period, no significant trend was detected.

In general, the high-resolution data from Finnpartner agreedreasonably with the pH/AT-based pCO2

w (Figs. 6a and 7). The slightbias towards higher pCO2

w, especially in the low-pCO2w range, was

attributed to the fact that the Finnpartner data do not referexactly to the BY15 position but represent the mean over severalmiles in the area. The consequences of the deviations between thetwo datasets for the annual CO2 flux budgets are discussed inSection 4.3. Another result of this comparison was the manydetails in the Finnpartner data that were missing from themonthly data. For example, in summer 2005, two summerminima were very well captured in the Finnpartner data, whilethis information was not detected in the calculated pCO2

w data. Tostatistically compare these two datasets, the calculated andmeasured pCO2

w, the mean error (ME, Eq. (5)), and the root

Fig. 6. Time series of calculated pCO2w from (a) the East Gotland Sea, (b) Bornholm Sea,

The grey thin line represents pCO2 in the atmosphere. In each panel, a trend line is dr

mean squared error (RMSE, Eq. (6)) were determined as follows:

ME¼1

N

XN

1

ðpCOw2 calc�pCOw

2 measÞ ð5Þ

RMSE¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1

N

XN

1

ðpCOw2 calc�pCOw

2 measÞ2

vuut ð6Þ

and (c) Kattegat. Panel (a) also includes pCO2w data from Finnpartner (open circles).

awn for the 1993–1998 period.

Page 7

Fig. 8. Air–sea exchange of CO2 in (a,b) the East Gotland Sea (c,d) the Bornholm Sea, and (e,f) the Kattegat. Panels (a), (c), and (e) show the annual net CO2 flux in

mol CO2 m–2 yr�1, grey bars show the flux with k according to Liss and Merlivat (1986), and the hatched bars show the flux with k according to Wanninkhof (1992). Black

bars are the total CO2 uptake and the white bars are the total release of CO2 each year. Panels (b), (d), and (f) show the daily flux of CO2 in mmol CO2 m–2 d–1. Flux is positive

from sea to air.

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–1521 1517

where N is the number of values. The time span betweencalculated and measured pCO2

w was less than 73 days. Thenegative ME, –1.3 matm, indicates that the calculated pCO2

w wason average lower than the measured pCO2

w. The RMSE of 49 matmreflects the size of the typical difference between calculated andmeasured pCO2

w, because of the combination of methods anddifference in sampling time. This highlights the risk whencomparing pCO2

w values from different sources.

3.2. The air–sea exchange of CO2

3.2.1. Seasonal variations

In accordance with the temporal pattern of pCO2w, there was a

clear seasonal cycle in the daily CO2 fluxes in the central BalticSea. In the central Baltic Sea, CO2 uptake was concentrated in thespring–summer period and CO2 release in the autumn–winterperiod (Fig. 8b and d). Such a seasonal cycle did not exist inKattegat (Fig. 8f), where CO2 was occasionally taken up from theatmosphere even in winter.

3.2.2. Inter-annual variations

Over the studied period, the central Baltic Sea and the Kattegatexperienced large inter-annual variations in the air–sea exchange

of CO2. Using the Wanninkhof (1992) parameterization of thetransfer velocity (Liss and Merlivat, 1986, values in parentheses),the three regions alternated between being annual net CO2 sinksand net CO2 sources of –4.2 (–2.5) to +5.2 (+3.1) mol CO2 m–2 y–1

(Fig. 8a, c, and e). On average throughout the 1994–2008 period,the East Gotland Sea was a source of 1.64(0.95) mol CO2 m–2 y–1,the Bornholm Sea was a source of 2.34 (1.38) mol CO2 m–2 y–1,while the Kattegat was a sink of –1.16 (–0.73) mol CO2 m–2 y–1.

The total CO2 uptake and release each year are also indicatedby the thin bars in Fig. 8a, c, and e. The total annual CO2 uptake inthe East Gotland Sea exceeded that in the Bornholm Sea and theCO2 release was almost the same. In Kattegat, both CO2 uptakeand release were of a smaller magnitude and the uptakedominated.

4. Discussion

4.1. The partial pressure of CO2

In winter in the central Baltic Sea, the water was super-saturated with CO2. This can be related to mineralization andvertical mixing. Mineralization exceeds production in winter, and

Page 8

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–15211518

when organic matter is mineralized, CO2 is released. The verticalmixing brings deeper CO2-enriched water to the surface. Thisprocess has been investigated by Schneider et al. (2000), whoidentified a net carbon flux across the halocline towards thesurface in winter and spring in the eastern Gotland Sea. InKattegat, the mixed layer is shallow, i.e., 20 m versus 60 m in thecentral Baltic Sea, and the vertical mixing does not reach thatdeep and hence does not bring that much CO2-enriched water tothe surface, which might explain the low winter pCO2

w in Kattegat.The wind speed (Fig. 9) was highest in winter, which also supportsthe vertical mixing in winter.

The rapid drop in pCO2w in spring coincided with the decrease

in phosphate (Fig. 10) and was thus coupled to the biologicalproduction. Biological production starts in spring when sunlightand nutrients are sufficient. In East Gotland Sea in 1993, the pCO2

w

was as low as 50 matm, and low values such as these have oftenbeen observed in mid-summer when nitrogen fixation activity hasbeen extremely high and when low wind speeds have impededvertical mixing (Schneider, unpublished data). During suchevents, extensive and compact mats of cyanobacteria areobserved at the sea surface. The high production of organicmatter together with the low buffer capacity of the Balticseawater then yields pCO2

w values as low as 50 matm, confined,however, to a shallow surface layer. These observations indicatethat the surface pCO2

w is controlled by biological processes andthat the increase in SST in spring and summer plays only aninferior role.

The importance of biological CO2 consumption/productionbecomes even more evident when the temperature effect isremoved from the pCO2

w data by normalizing to a constanttemperature (Takahashi et al., 2002)

pCO2wTmean ¼ pCO2

w� e½0:0423ðTmean�TobsÞ� ð7Þ

where Tmean is the mean, and Tobs is the observed temperature in1C for each region. The remaining variations in pCO2

w, largely

Fig. 9. Monthly mean wind speed in the Kattegat (black circles), Bo

Fig. 10. The phosphate concentration in the Kattegat (black circles), B

depend on changes in the CO2 content. The result indicates thatthe seasonal amplitude is larger for the pCO2

wTmean values (dashedlines in Fig. 11a–c). This means that in winter, if it were not for thetemperature effect that reduces pCO2

w, processes such asmineralization and mixing would have increased the pCO2

w evenmore. Similarly, biological production would have further reducedpCO2

w in summer if it were not for the temperature effect. It alsoseems that the temperature effect in the Kattegat is stronger,particularly in winter, and is more evident than in the centralBaltic Sea.

Several years in the central Baltic Sea had two distinct summerminima in pCO2

w. Schneider et al. (2006) explained the secondsummer minima in pCO2

w as resulting from a second seasonalproduction period due to nitrogen fixation, since the nitratesupply was exhausted at the time.

The stable thermocline in the central Baltic Sea implies thatproduction is separate from mineralization; these processes takeplace in different layers in the water column and there is clearseasonality. Kattegat, however, is shallower, mixes more fre-quently, and production and mineralization occur more often inthe same water layer. This explains the less pronouncedbiologically driven seasonality in the Kattegat. The varyingsalinities in Kattegat (Fig. 3c) also indicate that the drawdownof the pCO2

w by biological production was partly masked by theinput of various water masses and frequent deep mixing.

The standard deviation of the maximum absolute DpCO2 wasdetermined in spring/summer and in autumn/winter. In spring/summer, this value was 42 matm in the eastern Gotland Sea and50 matm in the Bornholm Sea. For the autumn/winter period, thevalue was 39 matm in the eastern Gotland Sea and 38 matm in theBornholm Sea. The standard deviation was hence, for both sites inthe central Baltic Sea, higher in summer than in winter. Thisimplies larger inter-annual variations in pCO2

w between summers,probably because of variations in biological production. This wassupported by the correlation between the maximum DpCO2 in

rnholm Sea (grey squares), and East Gotland Sea (open circles).

ornholm Sea (grey squares), and East Gotland Sea (open circles).

Page 9

Fig. 11. The temperature dependence of pCO2w in the East Gotland Sea, Bornholm Sea, and Kattegat. The black line is the calculated pCO2

w and the dashed line is pCO2wTmean

where the temperature effect has been removed.

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–1521 1519

summer and maximum phosphate concentration in winter(Fig. 12). The more available phosphate, the larger the biologicalproduction, and consequently, the larger the DpCO2. Two years inBornholm Sea, 1994 and 2005, however, did not support thiscorrelation. It was impossible to conduct a similar analysis inKattegat because there was no distinct summer or wintermaximum in DpCO2.

The peak of the winter pCO2w coincided with the maximum

salinity, indicating that the depth of the vertical mixing controlsthe increase in winter pCO2

w. However, there was no correlationbetween the magnitude of the maximum winter pCO2

w andsalinity.

The positive trends in pCO2w indicated in Fig. 6, in 1993–1998,

were probably associated with the deep water/sediment phos-phate dynamics. Before the 1993 major inflow, the deep waterhad experienced a long stagnation period during which largeamounts of phosphate were released from the sediment to thewater column (Stigebrandt and Gustafsson, 2007). After thisinflow, the phosphate content of the water column decreasedbecause of the storage in the sediment that was supported by theoxic conditions.

4.2. The air–sea exchange of CO2

The seasonality in the air–sea exchange of CO2 in thecentral Baltic Sea was a consequence of the seasonality in pCO2

w.

In Kattegat, where the seasonality in pCO2w was less distinct,

there was also less seasonality in the flux. The short-termvariations in the air–sea exchange were probably caused bythe changes in wind speed because we used daily wind speeddata.

The standard deviation of the outgassing fluxes in autumn/winter and input fluxes in spring/summer were determined in thecentral Baltic Sea. In autumn/winter, this value was 1.03 mol m–2 y–1

in the eastern Gotland Sea and 1.34 mol m–2 y–1 in the BornholmSea. In spring/summer, the value was 0.65 mol m–2 y–1 for theeastern Gotland Sea and 0.77 mol m–2 y–1 in the Bornholm Sea.Because the standard deviation of the outgassing fluxes was largerin autumn/winter, this indicated that the variability of the netannual fluxes was mainly controlled by autumn/winter condi-tions. Similarly, the standard deviation of the mean gas transfervelocity was determined. For the autumn/winter period, thestandard deviation was 1.48 cm h�1 for the eastern Gotland Seaand 1.69 cm h�1 in the Bornholm Sea. During spring/summercondition the same value was 1.13 cm h�1 in the eastern GotlandSea and 1.32 cm h�1 in the Bornholm Sea. This shows that theinter-annual variation in k was larger in winter than in summer,explaining why inter-annual variation in the flux was largest inwinter. Due to the missing seasonal cycle of the CO2 fluxes inKattegat, it was impossible to draw similar conclusions there. Theinter-annual variations in pCO2

w were largest in summer, but didhence not control the inter-annual variability of the air–seaexchange.

Page 10

Fig. 12. Maximum DpCO2 in summer versus maximum phosphate concentration

in winter in the Bornholm Sea (grey squares) and East Gotland Sea (open circles).

The regression lines are the grey thick line in the Bornholm Sea and black thin line

in the East Gotland Sea.

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–15211520

Thomas and Schneider (1999) estimated a net sink of–0.9 mol m–2 y–1 for the entire Baltic Proper, and a sink of –0.8and –1.7 mol m–2 y–1 for the Gotland and Bornholm seas,respectively. They based their results on the seasonal cycle ofpCO2

w at several locations, using measurements for 1994–1995together with a model using kW92 and climatological wind speed.Our estimates for the Gotland and Bornholm seas (average for thesame period, 1994–1995) were 0.9 and �0.4 mol m–2 y–1,respectively. The corresponding mean flux in the central BalticSea, 1994–1995, was a net source of 0.25 mol m–2 y–1. Hence, ourresults did not agree with those of Thomas and Schneider (1999),but were based on more data, which indicates the importanceof high data resolution. The Bornholm Sea was a larger netsource (2.34 mol m–2 y–1) than was the East Gotland Sea(1.64 mol m–2 y–1), probably because of the lower pCO2

w valuesin summer in the East Gotland Sea. To our knowledge, there areno previous studies of the annual net air–sea exchange of CO2 inthe Kattegat, so we cannot relate our findings for the Kattegat tothose of other studies.

Previous studies (e.g., Borges et al., 2006; Chen and Borges,2009) claim that continental shelves are net sinks, correspondingwell with our findings in Kattegat. The central Baltic Sea seems tofunction more as a CO2 source with large inter-annual variations.

4.3. Uncertainties

Uncertainties in the flux calculations (Eq. (2)) are mainlyassociated with the gas exchange transfer velocity, k, and the CO2

partial pressure difference, DpCO2, whereas the error of thesolubility constant, K0, is insignificant. Several bulk formula forthe parameterization of k as a function of wind speed exist andlead to results that may differ by more than a factor of two.Although the Wanninkhof (1992) formula is currently preferredby many researchers, it is still a matter of research to establish aparameterization that deserves general acceptance and includes arealistic uncertainty range. We therefore abstained from

estimating the flux uncertainty associated with k and consideredonly the uncertainty of the pCO2

w which affects the DpCO2. Usingthe equilibrium constants for the CO2 system (Millero et al., 2006;Weiss, 1974), the consequences of potential pH and AT measure-ment errors (70.015 and 72.5%, respectively) for the calculationof the pCO2

w were calculated. A mean variation of 74% wasobtained which being dependent on the magnitude of DpCO2 mayhave a considerable effect on the flux calculations. To demon-strate this, we subdivided the entire dataset in a CO2 uptake and aCO2 release period, which approximately corresponded to thespring/summer and the autumn/winter season, respectively. Themean pCO2

w for the uptake seasons varied between 240 matm(BY15), 281 matm (BY5), and 324 matm (AnE) and were associatedwith an uncertainty (4%) of 9.6, 11.2, and 12.9 matm, respectively.Relating these uncertainties to the mean DpCO2 (–130 matm atBY15, –89 matm at BY5 and –46 matm at AnE) yielded relativeerrors for the DpCO2 and thus for the fluxes of 7.4%, 12.6%, and27.8 %, respectively. The absolute values for the DpCO2 during theCO2 release seasons were smaller (87 matm at BY15, 84 matm atBY5, and only 23 matm at AnE) and at the same time the meanpCO2

w were higher. This resulted in larger uncertainties for theDpCO2. These were still moderate at BY15 (21.5%) and BY5(22.1%), but a value of 70.0% was obtained for the CO2 release fluxat the Kattegat station AnE which, however, constituted only asmall contribution to the annual budget (Fig. 8).

We also used flux calculations based on the quasi-continuousFinnpartner pCO2

w measurements in the vicinity of BY15 in orderto examine the quality of our flux estimates. Finnpartner datawere available from 2003 to 2005 (Fig. 5), but did not cover themonths December and January during which the surface watersact as a strong CO2 source. Therefore, the mean of the monthlyfluxes for the 3 years was slightly negative and amounted to –0.26 mol m�2. This CO2 uptake was by about 20% lower than themean flux that we obtained from the flux calculations based onthe monitoring data for the same months (–0.33 mol m�2). Inview of the different methods for the determination of the pCO2

w

and of the differences in the temporal resolution, we consider thisdiscrepancy as acceptable.

5. Conclusions

Inter-annual and seasonal variations in the air–sea CO2 balancehave been analyzed based on measurements from the centralBaltic Sea and the Kattegat. A 15 year period from 1994 to 2008has been studied based on monitoring data with a temporalresolution of one month and on direct CO2 measurements from acargo ship. The main conclusions were the following:

�

The net annual air–sea exchange of CO2 in the central BalticSea and the Kattegat varied both inter-annually and regionally.On average, for the 1994–2008 period, the East Gotland Seaand Bornholm Sea were sources of CO2 and the Kattegat was asink. � Inter-annual variations in pCO2

w minima were controlled by themaximum concentration of phosphate in winter.

� Inter-annual variations in the air–sea exchange of CO2 and the

gas transfer velocity were larger between winters thanbetween summers. This indicates that winter conditions werecontrolling inter-annual variability in the annual net flux.

� The large difference in sink/source function between the

central Baltic Sea and Kattegat was considered to depend ondifferent mixed layer depths and on different balancesbetween production and mineralization. In the central BalticSea, where the mixed layer depth is 60 m, CO2-enriched wateris mixed up to the surface in winter. In Kattegat, where the

Page 11

K. Wesslander et al. / Continental Shelf Research 30 (2010) 1511–1521 1521

mixed layer is shallower, i.e., 20 m, the mixing brings less CO2-enriched water to the surface. The central Baltic Sea alsoreceives large amounts of organic material from river waterinflow, which may lead to a heterotrophic system and thus to anet CO2 source. This is not the case in Kattegat, which is highlyinfluenced by oceanic conditions.

Acknowledgements

This work was part of the GEWEX/BALTEX and BONUS/Baltic-Cprogrammes and was financed by the University of Gothenburgand the Swedish Research Council under the G 600-335/2001contract. We wish to thank SMHI for contributing the monitoringdata and wind speed data and Dr. Agneta Fransson at theUniversity of Gothenburg for her valuable comments anddiscussion.

References

Algesten, G., Wikner, J., Sobek, S., Tranvik, L.J., Jansson, M., 2004. Seasonal variation ofCO2 saturation in the Gulf of Bothnia: indications of marine net heterotrophy.Global Biogeochem. Cycles 18, GB4021. doi:10.1029/2004GB002232.

Bergstrom, S., Carlsson, B., 1994. River runoff to the Baltic Sea: 1950–1990. AMBIO23 (4/5), 280–287.

Borges, A.V., Schiettecatte, L.-S., Abril, G., Delille, B., Gazeau, F., 2006. Carbondioxide in European coastal waters. Estuarine Coastal Shelf Sci 70, 375–387.

Caldeira, K., Wickett, M.E., 2003. Anthropogenic carbon and ocean pH. Nature 425,365.

Chen, C.-T.A., Borges, A.V., 2009. Reconciling opposing views on carbon cycling inthe coastal ocean: continental shelves as sinks and near-shore ecosystems assources of atmospheric CO2. Deep-Sea Res. II. doi:10.1016/j.dsr2.2009.01.001.

Dickson, A.G., Millero, F.J., 1987. A comparison of the equilibrium constants for thedissociation of carbonic acid in seawater media. Deep-Sea Res. 34 (10),1733–1743.

Dickson, A.G., Millero, F.J., 1989. A comparison of the equilibrium constants for thedissociation of carbonic acid in seawater media. Deep-Sea Res. 36, 983Corrigenda.

Dickson, A.G., 1990. Thermodynamics of the dissociation of boric acid in syntheticseawater from 273.15 to 318.15 K. Deep-Sea Res. 37, 755–766.

Fonselius, S., Valderrama, J., 2003. One hundred years of hydrographic measure-ments in the Baltic Sea. J. Sea Res. 49, 229–241.

Grasshoff, K., Kremling, K., Ehrhardt, M., 1999. Methods of Seawater Analysis 3rded. Wiley-VCH Verlag, Weinheim.

Hjalmarsson, S., Wesslander, K., Anderson, L.G., Omstedt, A., Perttila, M., Mintrop,L., 2008. Distribution, long-term development and mass balance calculation oftotal alkalinity in the Baltic Sea. Cont. Shelf Res. 28, 593–601.

Hogstrom, U., Sahlee, E., Drennan, W., Kahma, K.K., Smedman, A., Johansson, C.,Pettersson, H., Rutgersson, A., Tuomi, L., Zhang, F., Johansson, M., 2008.Momentum fluxes and wind gradients in the marine boundary layer: a multi-platform study. Boreal Env. Res. 13, 475–502.

ISO 10523: 1994. Water Quality—Determination of pH.Lewis, E., Wallace, D.W.R., 1998. Program Developed for CO2 System

Calculations. ORNL/CDIAC-105. Carbon Dioxide Information Analysis Center,Oak Ridge National Laboratory. U.S. Department of Energy, Oak Ridge,Tennessee.

Liss, P.S., Merlivat, L., 1986. Air–sea gas exchange rates: introduction andsynthesis. In: Buat-Menard, P. (Ed.), The Role of Air–Sea Exchange inGeochemical Cycling. Reidel, Doredrecht, pp. 113–127.

Mehrbach, C., Culberson, H., Hawley, J.E., Pytkowicz, R.M., 1973. Measurement ofthe apparent dissociation constants of carbonic acid in seawater at atmo-spheric pressure. Limnol. Oceanogr. 18 (6), 897–907.

Millero, F.J., 1995. Thermodynamics of the carbon dioxide system in the oceans.Geochem. et Cosmochem. Acta 59, 661–677.

Millero, F.J., Graham, T.B., Huang, F., Bustos-Serrano, H., Pierrot, D., 2006.Dissociation constants of carbonic acid in seawater as a function of salinityand temperature. Mar. Chem. 100, 80–94.

Morth, C.-M., Humborg, C., Eriksson, H., Danielsson, A, Medina, M.R., Lofgren, S.,Swaney, D.P., Rahm, L., 2007. Modeling riverine nutrient transport to the BalticSea: a large-scale approach. AMBIO 36, 124–133.

Omstedt, A., Axell, L.B., 2003. Modeling the variations of salinity and temperaturein the large Gulfs of the Baltic Sea. Cont. Shelf Res. 23, 265–294.

Omstedt, A., Elken, J., Lehmann, A., Piechura, J., 2004. Knowledge of the Baltic Seaphysics gained during the BALTEX and related programmes. Prog. Oceanogr.63, 1–28.

Omstedt, A., Chen, Y., Wesslander, K., 2005. A comparison between the ERA40 andthe SMHI gridded meteorological databases as applied to Baltic Sea modelling.Nord. Hydrol. 36 (4–5), 369–380.

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

Rutgersson, A., Norman, M., Schneider, B., Pettersson, H., Sahlee, E., 2008. Theannual cycle of carbon dioxide and parameters influencing the air–sea carbonexchange in the Baltic Proper. J. Mar. Syst. 74, 381–394.

Sabine, C.L., Feely, R.A., Gruber, N., Key, R.M., Lee, K., Bullister, J.L., Wanninkhof, R.,Wong, C.S., Wallace, D.W.R., Tilbrook, B., Millero, F.J., Peng, T.-H., Kozyr, A.,Ono, T., Rios, A.F., 2004. The oceanic sink for anthropogenic CO2. Science 305,367–371.

Schneider, B., Nagel, K., Struck, U., 2000. Carbon flux across the halocline in theeastern Gotland Sea. J. Mar. Syst. 25, 261–268.

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

Schneider, B., Kaitala, S., Raateoja, M., Sadkowiak, B., 2009. A nitrogen fixationestimate for the Baltic Sea based on continuous pCO2 measurements on a cargoship and total nitrogen data. Cont. Shelf Res. 29, 1535–1540.

Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M.,Miller, H.L., (Eds.), Climate change 2007: the physical science basis. Contribu-tion of Working Group I to the Fourth Assessment Report of the Intergovern-mental Panel on Climate Change. Cambridge University Press, Cambridge, UK,New York pp. 996.

Stigebrandt, A., Gustafsson, B.G., 2007. Improvement of Baltic Proper water qualityusing large-scale ecological engineering. AMBIO 36, 280–286.

Takahashi, T., Olafsson, J., Goddard, J.G., Chipman, D.W., Sutherland, S.C., 1993.Seasonal variation of CO2 and nutrients in the high latitude surface oceans: acomparative study. Global Biogeochem. Cycles 7, 843–878.

Takahashi, T., Sutherland, S.C., Sweeney, C., Poisson, A., Metzl, N., Tilbrook, B.,Bates, N., Wanninkhof, R., Feely, R., Sabine, C., Olafsson, J., Nojiri, Y., 2002.Global sea–air CO2 flux based on climatological surface ocean pCO2,and seasonal biological and temperature effects. Deep-Sea Res. II 49,1601–1622.

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

Thomas, H., Bozec, Y., Elkalay, K., de Baar, H.J.W., 2004. Enhanced open oceanstorage of CO2 from shelf sea pumping. Science 304, 1005–1008.

Tsunogai, S., Watanabe, S., Sato, T., 1999. Is there a ‘continental shelf pump’ for theabsorption of atmospheric CO2? Tellus 51B 701–712.

Wanninkhof, R., 1992. Relationship between wind speed and gas exchange overthe ocean. J. Geophys. Res. 97, 7373–7382.

Weiss, R.F., 1974. Carbon dioxide in water and seawater: the solubility of a non-ideal gas. Mar. Chem. 2, 203–215.

Weiss, R.F., Price, B.A., 1980. Nitrous oxide solubility in water and seawater. Mar.Chem. 8, 347–359.

Web references

For the xCO2measurements from the Polish coast, see:

http://gaw.kishou.go.jp/wdcgg/.2010-03-05For the geostrophic wind speeds, see:http://www.smhi.se/sgn0102/bhdc/.2010-03-05The monitoring data were available at:http://www.smhi.se/klimatdata/oceanografi/Marin-fysik-ke

mi-och-biologi/oceanografiskt-datacenter-1.3798.2010-03-05