Mechanisms controlling air-sea gas exchange in the Baltic Sea

(1)

Mechanisms controlling air-sea gas exchange in the Baltic Sea

Lucía Gutiérrez-Loza

(2)

Licentiate dissertation presented at Uppsala University to be publicly examined in Room Norrland I, Geocentrum, Villavägen 16, Uppsala, Wednesday, 10 June 2020 at 14:00.

The examination will be conducted in English. Opponent: Research Engineer John Prytherch (Department of Meteorology (MISU), Stockholm University).

Abstract

Gutiérrez-Loza, Lucía, 2020. Mechanisms controlling air-sea gas exchange in the Baltic Sea. 39 p. Uppsala.

Carbon plays a major role in physical and biogeochemical processes in the atmosphere, the biosphere, and the ocean. CO₂ and CH₄ are two of the most common carbon- containing compounds in the atmosphere, also recognized as major greenhouse gases.

The exchange of CO2 and CH4 between the ocean and the atmosphere is an essential part of the global carbon cycle. The exchange is controlled by the air–sea concentration gradient and by the efﬁciency of the transfer processes. The lack of knowledge about the forcing mechanisms affecting the exchange of these climate-relevant gases is a major source of uncertainty in the estimation of the global oceanic contributions. Quantifying and understanding the air–sea exchange processes is essential to constrain the estimates and to improve our knowledge about the current and future climate. In this thesis, the mechanisms controlling the air–sea gas exchange in the Baltic Sea are investigated.

The viability of micrometeorological techniques for CH4 monitoring in a coastal environment is evaluated. One year of semi-continuous measurements of air–sea CH₄ fluxes using eddy covariance measurements suggests that the method is useful for CH₄ flux estimations in marine environments. The measurements allow long-term monitoring at high frequency rates, thus, capturing the temporal variability of the flux. The region off Gotland is a net source of CH4, with both the air–sea concentration gradient and the wind as controlling mechanisms.

A sensitivity analysis of the gas transfer velocity is performed to evaluate the effect of the forcing mechanisms controlling the air–sea CO₂ exchange in the Baltic Sea. This analysis shows that the spatio-temporal variability of CO2 ﬂuxes is strongly modulated by water-side convection, precipitation, and surfactants. The effect of these factors is relevant both at regional and global scales, as they are not included in the current budget estimates.

Keywords: Air-sea gas exchange, Baltic Sea, CO2 flux, methane flux

Lucía Gutiérrez-Loza, Department of Earth Sciences, Program for Air, Water and Land- scape Sciences; Meteorology, Geocentrum, Villavägen 16, 752 36 Uppsala.

urn:nbn:se:uu:diva-409744 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-409744)

(3)

To Héctor Miguel

(4)

(5)

List of papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Gutiérrez-Loza, L., Wallin, M.B., Sahlée, E., Nilsson, E., Bange, H.W., Kock, A., Rutgersson, A. (2019) Measurement of air–sea methane fluxes in the Baltic Sea using the eddy covariance method.

Front. Earth Sci., 7:93. doi:10.3389/feart.2019.00093

II Gutiérrez-Loza, L., Rutgersson, A., Wallin, M.B., Sahlée, E., Shutler, J.D., Holding, T., Rehder, G. (2020) Air–sea CO₂exchange in the Baltic Sea—a sensitivity analysis of the gas transfer velocity.

Manuscript.

Reprints were made with permission from the publishers.

For Paper I the author participated in performing the measurements and was responsible of most of the data analysis. In Paper II the author had the main responsibility of running the FluxEngine toolbox and analyse the results. The author had the main responsibility of writing Papers I and II.

(6)

(7)

1. Introduction

Carbon plays a major role in physical and biogeochemical processes in the atmosphere, the biosphere, and the ocean. In the atmosphere, carbon is present—

among other compounds—in the form of carbon dioxide (CO₂) and methane (CH4), which are recognized to be the major greenhouse gases. Since pre- industrial times, the concentration of these climate-relevant gases has significantly increased due to human-related activities like fossil fuel burning and the change in land use. In the last 60 years, the average atmospheric CO₂ concentration has been steadily increasing from ∼300 ppm in the 1950s up to today’s value of 412.5 ppm (Dlugokencky et al., 2020). In a similar way, the global average of CH₄ atmospheric concentration has more than doubled, reaching values of over 1800 ppb (WDCGG, 2012). The anthropogenic emissions of atmospheric greenhouse gases are an additional contribution to the natural variations which also have a relevant role in the global carbon cycle at several spatial and temporal scales.

The oceans are a net sink of atmospheric carbon taking up around 24%

of the global anthropogenic CO2emissions in the last decade (Friedlingstein et al., 2019). Estimates of the global oceanic carbon uptake rates range between −1.0 and −3.2 PgC/year (Rhein et al., 2013; Friedlingstein et al., 2019).

The bulk methods used for the global air–sea CO₂ flux estimates are associated with large uncertainties (see Woolf et al., 2019) and do not adequately account for the temporal and spatial variability of the fluxes. On the other hand, the oceans are suggested to be a net source of methane and emit 11-18 TgCH4/year to the atmosphere (Bange et al., 1994; Reeburgh, 2007), but this is highly uncertain as air–sea CH₄ flux has been poorly studied. In order to address the uncertainties on the global estimates of the CO2and CH4budgets, studies resolving the local and regional air–sea gas exchange processes are necessary. Furthermore, an accurate assessment of the role of the oceans on CO2 and CH4 budgets is still missing. These contributions to the knowledge about oceanic carbon cycle may represent a better understanding of the current climate and its response to increasing greenhouse gases.

Coastal oceans have been recently recognized to be an important component of the global carbon cycle (IPCC, 2013). They are responsible for up to 30% of the oceanic primary production, 30-50% of the inorganic carbon, and up to 80% of the carbon burial in sediments (Gattuso et al., 1998; Jahnke, 2010). These biologically active regions also contribute with 75% of the global oceanic CH4 emissions (Bange et al., 1994). Air–sea gas fluxes in these het- erogeneous environments are modulated by complex interactions of physical

(10)

and biogeochemical processes which are poorly understood, and therefore, very seldom included in the global carbon budgets. Studies characterizing the global coastal oceans, and their role as net sinks or sources of CO₂ to the atmosphere have shown that there is a significant latitudinal variability (Borges et al., 2005; Cai et al., 2006; Chen et al., 2013). The estimates, however, do not account for the seasonal or the inter-annual variability of the fluxes. Al- though the number of studies has increased considerably in the last decades, the forcing mechanisms modulating the air–sea CO2fluxes in coastal areas and marginal seas are still poorly understood. Moreover, the role of these regions as net sinks or sources of atmospheric carbon is uncertain.

The largest uncertainties of air–sea gas exchange estimates are associated with the paucity of the data, and with uncertainties in the gas transfer velocity parameterizations. Long-term monitoring of the fluxes and an improved knowledge about the mechanisms controlling the gas transfer efficiency at local and regional scales are necessary to constrain these uncertainties. The Baltic Sea—a marginal sea on relatively high latitudes—has been subject of several studies addressing the carbon system elements and the air–sea interactions (Bange et al., 1994; Omstedt et al., 2009; Rutgersson et al., 2009;

Thomas et al., 1999; Wesslander et al., 2010). The effect of water-side convection (Rutgersson et al., 2010; Rutgersson et al., 2011), coastal upwelling (Norman et al., 2013a), and the other relevant parameters influencing the gas exchange (Rutgersson et al., 2008) have been studied in the Baltic Sea. Fur- thermore, studies highlighting the potential of using land-based stations (Nor- man et al., 2012; Rutgersson et al., 2020), remote sensing (Parard et al., 2017), and models (Norman et al., 2013b) for air–sea interaction studies have been carried out in the context of the Baltic Sea. The knowledge acquired from these regional studies, can contribute to the general understanding of air–sea gas exchange processes in the marginal seas and coastal regions, in addition to their relevance from the global perspective.

The aim of this work is to contribute to a better understanding of the air–

sea gas exchange processes. We explore the use of micrometeorological techniques for long-term CH4monitoring and assess the forcing mechanisms controlling the air–sea gas exchange in the Baltic Sea.

• In Paper I we investigate the viability of using micrometerological techniques (i.e. eddy covariance) for continuous measurements of air–sea CH4

fluxes in the coastal region. The continuous monitoring of CH4fluxes using high-frequency measurements potentially will allow to capture the temporal variability and forcing processes of air–sea CH₄in coastal environments.

• In Paper II we present a sensitivity analysis of the gas transfer velocity to evaluate the effect of the forcing mechanisms on air–sea CO₂ exchange in the Baltic Sea.

(11)

2. Theory

2.1 Air–sea gas exchange

The air–sea gas exchange of slightly soluble gases (e.g. CO2 and CH4) occurs due to a concentration difference between the top and the bottom of the oceanic boundary layer. In addition to the thermodynamic forcing (i.e. the concentration gradient), the turbulent processes near the ocean surface modu- late the efficiency of the transport across the air-sea interface. These processes are represented by the so-called gas transfer velocity (k). Assuming that the concentration at the water surface is in chemical equilibrium with the air just above the interface, is possible to express the flux (F) as

F= k(Cw− αCa), (2.1)

where the gradient is given by the concentrations of the gas in the bulk seawater (Cw) and in the air (Ca), and α is the dimensionless solubility coefficient.

By convention, positive fluxes refer to transport from the ocean to the atmosphere.

Air–sea gas exchange estimations can be performed from Equation 2.1, or by using similar equations where the concentration gradient is expressed in terms of the partial pressure or the fugacity of the gas. The use of such bulk methods, although convenient, involves large uncertainties mostly associated with the parameterization of the gas transfer velocity. Furthermore, they usually lack the capability of accurately account for the temporal and spatial variability of the flux.

2.2 The gas transfer velocity

The gas transfer velocity represents the efficiency of the transfer processes across the interface. From a two-layer model approach (Liss et al., 1974), k is inversely proportional to the total resistance exerted by the diffusive sub-layers to the gas transfer,

k⁻¹= Rw+ Ra (2.2)

where Rw and Ra are the water and air resistances, respectively. Whether the flux of a gas will be controlled by the air-side or water-side resistance depends on the characteristics of the gas. Slightly soluble gases (like CO2 and CH4)

(12)

are mainly controlled by processes occurring in the water-side of the interface, while the flux of highly soluble gases (e.g. water vapour) is controlled by air- side processes. Thus, turbulent processes in the oceanic boundary layer are relevant for air–sea exchange of CO2and CH4.

The water-side gas transfer velocity can be expressed in mathematical terms as

kw= aSc⁻ⁿf(Q, L, ν), (2.3)

where a is a proportionality constant. The Schmidt number (Sc = ν/D) de- scribes the extent of the diffusion (in terms of the diffusion coefficient, D) across the interface in comparison to kinematic viscosity (ν), and n is Schmidt number exponent (Esters et al., 2017; Jähne et al., 1984). The function, f , de- scribes the turbulent characteristics of the aqueous boundary layer in terms of a velocity scale (Q), a length scale (L), and the kinematic viscosity of the water.

In practice, it is convenient to express the gas transfer velocity in a polynomial form (Woolf et al., 2019). The polynomial coefficients are determined experimentally, most often as a wind speed function (Ho et al., 2006; McGillis et al., 2001; Nightingale et al., 2000; Wanninkhof, 1992; Wanninkhof et al., 2009). Wind is one of the primary energy inputs to the surface ocean, and one of the causes of its turbulent behaviour. It is, directly or indirectly, the driving force of most of the physical processes near the sea surface. Hence, the wind is a good proxy to represent the turbulent processes that control the air–sea gas exchange. However, neglecting other forcing mechanisms may lead to large uncertainties in the flux estimates, specially at local and regional scales.

(13)

3. The study site

3.1 The Baltic Sea

The Baltic Sea is a semi-enclosed sea in Northern Europe (Figure 3.1), it is considered one of the largest extents of brackish waters in the world. The basin stretches from 54^◦ to 66^◦ and has a total surface area of 420,000 km². The total catchment area is more than four times larger than the surface area of the Baltic Sea, covering from the temperate South to the boreal North (Räike et al., 2019). The diversity of the ecosystems in the Baltic Sea is highly influ- enced by large atmospheric circulation and hydrological processes. Moreover, restricted water exchange with the open ocean in the South and high fresh water input in the North promote estuarine-like conditions. These features play a fundamental role on the spatio-temporal variability of physical properties, biota, and biogeochemical processes; and therefore, in the variability of the carbon system elements in the Baltic Sea. In Paper II we discuss the seasonal and inter-annual variability of CO2air–sea fluxes induced by different forcing mechanisms in four sub-basins of the Baltic Sea.

Longitude Latitude

5°E

10°E 15°E 20°E 25°E 30°E 51°N

54°N 57°N

60°N 63°N

66°N

5°E

10°E 15°E 20°E 25°E 30°E 51°N

54°N 57°N

60°N 63°N

66°N

Figure 3.1. Map of the Baltic Sea. The red mark indicates the location of the Öster- garnsholm station.

(14)

3.2 The Östergarnsholm station

The Östergarnsholm measurement site is located in a flat and small island in the central Baltic Sea a few kilometers off the Gotland island (red mark in Fig- ure 3.1). The station has a 30-m land-based meteorological tower suitable for studying the marine atmospheric boundary layer and air–sea interaction processes (Rutgersson et al., 2020). The tower is equipped with high-frequency EC systems at two levels (9 and 25 m from the tower base), and with slow- response instrumentation at 5 levels to measure the mean profile of meteorological variables. In addition to the atmospheric measurements, water-side data is obtained at 4-m depth in a buoy located 1 km to the south-east of the tower. The buoy is instrumented to measure CO₂ partial pressure (pCO^w₂), water temperature, conductivity, depth, and oxygen concentration.

The measurements at Östergarnsholm station are representative of open-sea conditions for wind directions between 80^◦< WD < 220^◦. For these wind directions the wave field is considered to be undisturbed, the limited water depth and the coast have no noticeable effect on the measured quantities (Högström et al., 2008). For wind directions between 50^◦< WD < 80^◦and 220^◦< WD <

295^◦, the physical and biogeochemical properties may be affected by the coast (Rutgersson et al., 2020). Thus, the measurements from those wind directions represent coastal conditions. Measurements from Östergarnsholm station are used in Paper I for air–sea CH4flux estimations.

Figure 3.2. Photo of the Östergarnsholm station in the Baltic Sea. The 30-m tower is the rightmost construction.

(15)

4. Methods

4.1 The eddy covariance method

The eddy covariance (EC) method allows direct estimation of the magnitude and direction of turbulent fluxes from high-frequency data (Aubinet et al., 2012). The EC method was first developed in the mid 1900s (Foken et al., 2012) and has become one of most robust techniques for flux estimations. Mi- crometeorological techniques, including EC, have been widely used in terres- trial applications for momentum, energy, and mass flux estimations (Baldoc- chi et al., 2001). In marine environments, the use of these techniques has been rapidly increasing during the last decades mostly for CO₂ and water vapour flux estimations. In Paper I, we use the EC method to estimate the air–sea CH4fluxes at the Östergarnsholm station, presenting one of the first results of CH₄fluxes from EC in a marine environment.

From a theoretical perspective, the mean vertical turbulent flux (F) of any scalar can be derived from the conservation equation if i) the surface is hor- izontally uniform, ii) no sources or sinks of the material are present in the atmosphere above the surface, and iii) steady-state conditions can be assumed for both the wind and the scalar (Baldocchi et al., 1988). When these assumptions are valid, the air–sea gas exchange can be described using the eddy covariance as:

F= ρaw⁰s⁰, (4.1)

where ρais the density of dry air, w is the vertical component of the wind velocity, and s is—in this case—the gas mixing ratio (dry mole fraction) (Sahlée et al., 2008). The overbar represents the temporal average (usually 30 min) and the primes (⁰) represent the turbulent fluctuations estimated through a Reynold’s decomposition:

x= x + x⁰, (4.2)

here x is the measured signal, x represents the time-averaged value, and x⁰ represents the fluctuating part used for the flux calculations.

EC is one of the most direct methods for gas flux estimations. There are, however, several drawbacks from the practical perspective when using the EC methodology. On the one hand, it is unlikely that the assumptions of steady- state conditions and spatial homogeneity are fully met (Moncrieff et al., 2004).

On the other hand, instrumental limitations and technical issues (e.g. power

(16)

supply, site accessibility, and instrumental set up) may compromise the high quality of the data. Rigorous data-processing procedures must be followed to overcome the limitations and fulfill the theoretical requirements of the method.

4.1.1 Data treatment and quality control

In Paper I, air–sea CH₄ fluxes are estimated using the EC method for the period between October 1st, 2017 and September 30th, 2018. The high- frequency measurements of wind speed and atmospheric CH4 mole fractions from the Östergarnsholm station are subjected to a rigorous data selection and quality-control procedures prior to the flux estimations (Table 4.1).

Table 4.1. Summary of data treatment and quality control procedures for air–sea CH₄ flux estimations using the EC method.

Procedure Method/Value Reference

Coordinate rotation Double-rotation method Kaimal et al. (1994) Filtering Non-linear median filter Starkenburg et al. (2016)

De-trending Linear fit Moncrieff et al. (2004)

Averaging Block averaging^? Moncrieff et al. (2004) Minimum CH₄conc. 1800 ppb Baldocchi et al. (2012) Wind direction selection 50^◦< WD < 295^◦ Rutgersson et al. (2020) Density correction Webb-Pearman-Leuning

correction

Webb et al. (1980)

Spectroscopic correction − McDermitt et al. (2011) Detection threshold ±4 nmol m⁻²s⁻¹ Baldocchi et al. (2012)

?Averages were performed over 30-minute periods.

In addition to the procedures mentioned in Table 4.1, some statistical parameters are used to ensure the homogeneity of the data and to avoid outliers.

Data is excluded if the standard deviation of the 30-min averages of CH₄ dry mole fraction exceeds 35 ppb (Baldocchi et al., 2012), and when its 4th-order moment is higher than 100 ppb⁴. The skewness and kurtosis must range between −2 - 2 and 1 - 8, respectively, values outside these ranges are filtered out (Vickers et al., 1997). The relative signal strength (RSSI) is an indicator of the state of the optical mirrors, and therefore, of the quality of the data. Data is discarded if RSSI< 10% (Podgrajsek et al., 2016).

4.2 The MEMENTO database

(17)

and CH4 data (Kock et al., 2015). The data corresponds to in-situ measurements taken worldwide over the past 50 years. In Paper I, the measurements from the Östergarnsholm station are compared to CH₄ data reported in the MEMENTO database (Table 1 in Paper I). The data includes dissolved gas concentrations and corresponding atmospheric concentration values of CH₄in the Baltic Sea collected between 1986 and 2015. To the best of our knowledge, the MEMENTO database contained all the CH₄ data available in the Baltic Sea at the time of publication. Therefore, it is used as the only means to validate the Östergarnsholm measurements and flux estimates from EC.

4.3 The FluxEngine toolbox

The FluxEngine toolbox is an open source software for air–sea gas exchange calculations. A detailed description of the toolbox can be found in Shutler et al. (2016) and Holding et al. (2019). In Paper II, we use FluxEngine to calculate the air–sea CO₂ flux in the Baltic Sea and its four sub-basins. The fluxes are calculated following the rapid model approach (Woolf et al., 2016):

F= k(αwfw− αsfa), (4.3) where k is the gas transfer velocity, αw and αsare the solubilities of the gas in the bulk water and right at the surface, respectively, and fw and fa are the corresponding fugacities.

The impact of relevant mechanisms affecting the air–sea CO₂ exchange is evaluated in Paper II by using different formulations of the gas transfer velocity in Equation 4.3. The base-case scenario (N00) considers solely a wind-dependent formulation, while the other cases include the effect of water-side convection (N00+conv), rain (N00+rain), or surfactant suppression (N00+surf). Another case includes the effect of all the afore mentioned parameters (N00+all), and a final case that also includes water-side convection (RS10) serves for testing a parameterization developed using data from the Baltic Sea. A summary of the six parameterizations used in the study is shown in Table 4.2.

4.3.1 Input data

The FluxEngine allows the incorporation of different data sources (i.e. Earth observation, model, and in-situ data) for the air–sea flux calculations. In Pa- per II we use three years of data (2009-2011) from different sources (Table 4.3). The input datasets are time-averaged to get monthly means, and spatially averaged to match the spatial grid covering from 52^◦to 66^◦N and from 10^◦to 32^◦E with a resolution of 1/24^◦(equivalent to a 4 x 4 km grid at the equator).

(18)

Table 4.2. Summary of the gas transfer velocity parameterizations (modified from Paper II).

Name Description Reference

N00 k= k_uwhere k_u= 0.222U₁₀² + 0.333U10 Nightingale et al. (2000) N00+conv k= k_u+ k_cwhere k_c= 3022w − 20 Rutgersson et al. (2010) N00+rain k= k_u+ k_rwhere Ashton et al. (2016);

k_r= 0.929 + 0.679R_n− 0.0015R²_n

N00+surf k= k_uRwhere Pereira et al. (2018)

R= 1 − (0.0046 ∗ Tskin)^2.5673

N00+all k= (k_u+ k_c+ k_r)R −

RS10 k= k_u+ k_cwhere k_u= 0.24U₁₀² Rutgersson et al. (2010) Nomenclature: w is the convective velocity scale; R_nis the precipitation rate; R is the surfactant suppression factor.

?All cases use k_ufrom the base-case scenario (N00) except case RS10.

??The transfer velocity (k) in all cases is normalized using the Schmidt number (Sc).

Table 4.3. Data sources (see Section 3.2 in Paper II for details).

Name Parameter Source

x_a Atmospheric CO₂concentration Calculated following Rutgersson et al. (2010) and Norman et al.

(2013b)

pCO₂ Partial pressure CO₂in seawater From Parard et al. (2016) T_pco2 Temperature at pCO₂measurement From Parard et al. (2016)

U₁₀ Wind speed Reanalysis data from NEWA^?

T_skin Skin temperature Data from NEWA^? P Atmospheric pressure From NCEP reanalysis^??

R_n Precipitation rate From CMAP^??

S Sea surface salinity Renalysis product provided by CMEMS^???

MLD Mixed layer depth Renalysis product provided by CMEMS^???

ice Fraction of sea ice coverage Renalysis product provided by CMEMS^???

? New European Wind Atlas (NEWA) available at https://map.

neweuropeanwindatlas.

??Provided by NOAA/OAR/ESRL PSD available athttps://www.esrl.noaa.

gov/psd/.

??? Copernicus Marine Environment Monitoring System (CMEMS) available at http://marine.copernicus.eu/services-portfolio/

(19)

5. Results

5.1 Air–sea CH

₄

fluxes in the Baltic Sea

Estimates of the total oceanic CH₄contribution to the atmosphere are limited by the scarce CH4 data availability in marine environments. In order to understand the net air–sea CH₄exchange, measurements capturing the temporal and spatial resolution of the exchange processes are necessary. In Paper I, we present one year of semi-continuous measurements of air–sea CH₄ fluxes at the Östergarnsholm station. The results suggest that micrometeorological methods, in particular, the eddy covariance (EC) method can be used for direct estimation and monitoring of air–sea CH₄fluxes in marine environments.

The results are validated with CH4 in-situdata in the Baltic Sea available in the MEMENTO database (Section 4.2). The data reported in the MEMENTO database corresponds to individual studies performed between 1986 and 2015 in the Baltic Sea (Table 1 in Paper I).

The atmospheric concentration of CH4in the Central Baltic Sea has a seasonal variability with lower values during summer, probably caused by a higher consumption rate of CH₄in the atmosphere due to OH radicals. The seasonal variability is captured by the measurements at Östergarnsholm station and sup- ported by the data in the MEMENTO database (Figure 5.1). The magnitude of the values observed from the Östergarnsholm measurements (2004.6 - 2217.7 ppb) is significantly higher than those reported previously in the MEMENTO database over the 30-year period (1860.0 - 1905.0 ppb). The measurements at Östergarnsholm seem to be consistent with the global trends that suggest an increase of ∼200 ppb in the atmospheric CH₄ concentrations since 1985.

Although we consider the observations at the Östergarnsholm station to be reliable, we highlight the importance of implementing more data sources pro- viding continuous and high-quality measurements of atmospheric CH4 concentrations in the Baltic Sea.

The air–sea CH4 fluxes are estimated using the EC method over 30-min averages and following strict data selection criteria and high-quality control procedures (Section 4.1.1). The 30-min values present a large scatter with both positive (upward) and negative (downward) CH4 fluxes over the study period (Figure 3d in Paper I). This large scatter in the data exposes the turbulent behaviour of the processes involved in the gas exchange, as well as the importance of high-frequency measurements to capture the temporal variability of the flux. The positive monthly means suggest that the study area is a net source of CH4to the atmosphere at a rate of 3.7 - 36.1 nmol m⁻²s⁻¹.

(20)

Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Month

1850 1900 1950

(b) Atmospheric CH4(ppb)

MEMENTO (1986-2015) 2000

2100 2200

(a) Atmospheric CH4(ppb)

This study (2017-2018)

Figure 5.1. Monthly means of atmospheric CH₄ dry mole fraction from a) Öster- garnsholm station during 2017-2018, and b) MEMENTO database obtained between 1986 and 2015 (from Paper I).

One of the advantages of using the EC method for air–sea flux calculations is that it does not rely on any bulk parameterizaton (i.e. Equation 2.1). How- ever, using EC and bulk equations simultaneously can provide useful informa- tion about the gas transfer processes. In order to understand the mechanisms driving the air–sea CH4exchange, we analyse the impact of the concentration difference (∆C) on the air–sea CH4flux by comparing the annual cycles (Fig- ure 5.2). The concentration gradient is mostly modulated by changes in the water-side concentrations while the atmospheric values remain fairly constant (Figure 5.2a). During the summer months, the increase in the air–sea CH4

flux is consistent with the increase in ∆C, indicating that the fluxes are mostly driven by the thermodynamic forcing. In contrast, the large CH4 fluxes observed during the winter months are not explained by the ∆C values that only show a small increase. Stronger mean wind speed values and high-wind speed events observed during the winter seem to be driving the flux instead (Figure 3a in Paper I).

Wind speed (U ) is one of the parameters often used to describe the efficiency of the transport processes across the air-sea interface. A large correlation between the air–sea flux and the wind speed is, therefore, expected. In this study, we observe that such correlation only occurs under high wind-speed

(21)

Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Month

-10 0 10 20 30 40

FCH4(nmol m-2s-1) (b)

(ppb)

1500 2000 2500 3000 3500

0 500 1000 1500

CH4(ppb)

C = sea - atm Seawater

Atmosphere

(a)

C

Figure 5.2. Annual cycle of a) seawater and atmospheric CH₄mole fractions (prin- cipal y-axis) and concentration gradient (secondary y-axis), and b) air–sea CH₄flux from EC measurements (from Paper I).

velocity (k). Such comparison was not performed in this study due to the lack of continuous water-side measurements for k estimation using, for example, Equation 2.1.

The analysis of one of the high-wind speed event during January 14-16, 2018, shows that the relationship between the wind speed and the CH₄flux is not linear (Figure 5.3). While the wind speed is steadily increasing from the beginning of the event and for almost 48 hours, it takes about 24 hours for the air–sea CH4flux to show an increase. The delay in the response of the CH4

flux and the corresponding atmospheric CH4concentration, suggest that there is a threshold value of the wind speed below which the wind-generated turbulence does not create enough mixing in the upper ocean. The threshold might be either directly dependent of the wind speed, in which case a minimum wind speed is necessary; or alternatively, the threshold is time-dependent for which case the turbulence develops under the effect of the wind and over time until it reaches the point when the mixing is sufficient to enhance the transport across the interface. Either way, this is a first approach to evaluate the forcing mechanism of air–sea CH4 fluxes and further analysis of the turbulent processes modulating the efficiency of the CH4is necessary.

(22)

08:00 16:00 00:00 08:00 16:00 00:00 08:00 4

6 8 10 12 14 16 18 20

Wind speed (m s-1)

(a)

08:00 16:00 00:00 08:00 16:00 00:00 08:00 0

25 50 75 100 125 150 175 200

FCH4(nmol m-2s-1)

(b)

2000 2100 2200

Atmospheric CH4 (ppb)

Figure 5.3. a) Wind speed, and b) air–sea CH4₄fluxes over a 48-h period during a high wind-speed event (from Paper I).

5.2 Sensitivity analysis of the gas transfer velocity

Wind-based parameterizations used for global estimates of air–sea gas exchange are seldom suitable for flux estimations at regional scales. In particular, wind-based estimates entail large uncertainties when used in heteroge- neous environments such as coastal oceans and marginal seas. Using the wind speed as a proxy of the turbulent processes controlling the efficiency of the gas transfer can be advantageous. However, understanding how the relevant mechanisms directly affect the gas exchange is necessary to adequately estimate the carbon budgets at regional scales. In Paper II we use six different gas transfer velocity parameterizations (Table 4.2) to evaluate the effect of precipitation, water-side convection, and surfactants on air–sea CO2exchange in the Baltic Sea. The air–sea fluxes are calculated using a bulk method following a rapid model approach (Equation 4.3).

The annual cycle of the air–sea CO2 flux obtained as the average over a three-year period (2009 to 2011) shows that the Baltic Sea is a net source of CO2 to the atmosphere during the winter months, while it acts as a net sink during the summer (Figure 5.4). Similar to what we found for air–sea CH4 fluxes in Paper I, the seasonal cycle of CO2 flux is mostly modulated by the concentration gradient (expressed as ∆pCO2in Figure 2f in Paper II).

The seasonality of ∆pCO₂ is mostly caused by biological processes in the ocean, photosynthesis in summer and respiration in winter, controlling the consumption and production of CO₂ in the seawater. The atmospheric CO₂ concentrations have a fairly low variability in comparison to the changes in the water-side concentrations.

(23)

Both upward (positive) and downward (negative) fluxes are affected by these processes. Water-side convection is significant during the winter months, when convective processes occur due to the heat loss at the ocean surface.

The effect of water-side convection is noticeable as an increase in the upward fluxes, especially in the Gulf of Bothnia and the Central Basin. During summer, precipitation and surfactant suppression are competing mechanisms with precipitation increasing the downward (negative) flux, and surfactants sup- pressing the flux.

J F M A M J J A S O N D

Month -0.3

-0.2 -0.1 0 0.1 0.2 0.3

CO2 Flux [gC m-2d-1]

N00 N00+conv N00+rain N00+surf N00+all RS10

Figure 5.4.Average annual cycle of air–sea CO₂fluxes calculated using six gas transfer velocity parameterizations (from Paper II).

The annual mean air-sea CO2 flux indicates that the Baltic Sea is nearly at equilibrium, ranging between 0.0 and −0.02 gC m⁻²s⁻¹ among the six cases. Even though including the different driving mechanisms represents only a small change in the mean annual flux, it has relevant implications in the seasonal behaviour (Figure 5.4), and represents significant changes in the total contributions in terms of the net annual flux (Figure 5.5a). The inter-annual variability is also modulated by the driving mechanisms (Figure 5.5b). In all cases, the water-side convection seems to be the parameter affecting the fluxes to the major extent, and even change the sign of the net flux from negative (sink) to positive (source). Example of this is the year of 2010, when positive net fluxes are observed when water-side convection is taken into account.

The spatial distribution is evaluated by analysing the air–sea CO₂fluxes in four sub-basins of the Baltic Sea: Gulf of Bothnia, Gulf of Finland, Central Basin, and Southern Basin (Figure 1 in Paper II). The mean annual net CO₂ flux in the four sub-basins shows that both the flux and the effect of the driving mechanisms present a spatial distribution (Figure 5.6). The variability is associated to the characteristics of each individual basin. In the northern regions, strong cooling of the water surface occurs during the winter months, there-

(24)

-1.5 -1 -0.5 0

Net flux [TgC y-1]

(a) -73.3%

+27.6%

-0.7%

-51.4%

-99.8%

N00

N00+conv N00+rain N00+surf N00+all RS10

-2 -1.5 -1 -0.5 0 0.5 1

Net flux [TgC y-1]

2009 2010 2011

(b)

Figure 5.5. a) Average net CO₂flux in the Baltic Sea, and b) Annual net CO₂flux in the Baltic Sea for the years 2009, 2010, and 2011. The numbers in the top panel represent the percentage of increase/reduction in the net flux with respect to the base case scenario (from Paper II).

fore, water-side convection seems to be the most relevant parameter affecting the net CO₂ fluxes in the Gulf of Bothnia and the Central Basin. The effect of precipitation is also stronger in the Gulf of Bothnia and the Central Basin, which are the largest basins. Thus, the effect of precipitation may be associated with the surface extent more than with the latitudinal variability. Surfac- tant suppression seems to be less significant than water-side convection and precipitation, but its effect is noticeable in the four sub-basins. The Southern Basin is moderately affected by all the parameters, while, the Gulf of Finland is the least affected basin indicating that wind-based parameterizations might be suitable for the region.

The differences in the air–sea CO2fluxes relative to the base-case scenario are directly linked to the gas transfer velocity parameterization. The annual cycle of the gas transfer velocities is strongly modulated by the wind speed patterns, with larger values during the winter and lower values during spring and summer (Figure 5.7). However, we also observe differences between the six different cases, suggesting that the wind-speed is not the only relevant mechanism modulating the exchange, as was also observed from the flux estimates. Convective processes seem to be particularly relevant during winter

(25)

N00 N00+conv N00+rain N00+surf RS10 N00+all

GB GF CB SB

-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1

Net flux [TgC y-1]

Figure 5.6. Net CO₂ flux in the four sub-basins of the Baltic Sea: Gulf of Both- nia (GB), Gulf of Finland (GF), Central Basin (CB), and Southern Basin (SB) (from Paper II).

hancing the upward fluxes that occur in the winter months. Surfactants have an effect during summer when the surface water temperatures are high and strong biological activity occurs. The effect of surfactants is perceived by the low values of k from June to November. The result is the restriction of the (downward) fluxes during the summer months and, to a lesser extent, in the (upward) fluxes towards the end of the year. The effect of precipitation is perceivable throughout the year which results in the increased magnitude of both upward (winter) and downward (summer) fluxes in comparison to the base-case scenario. The results of the six gas transfer velocity formulations used in this study lay within the spread range of commonly-used gas transfer velocity parameterization (shaded area in Figure 5.7). This indicates that the uncertainty associated with the selection of the wind-based parameterization may be larger than the effect of neglecting the other forcing mechanisms. It is important to notice, however, that the effect of the mechanisms evaluated here is an additional term to consider in the uncertainty of the flux estimations.

(26)

J F M A M J J A S O N D Month

5 10 15 20 25

k [cm h-1]

Figure 5.7. Monthly means of the gas transfer velocity in the Baltic Sea using six different gas transfer velocity parameterizations. The gray dashed-line represents the mean gas transfer velocity from a set of wind-based parameterizations, the shaded area is the standard deviation (from Paper II).

(27)

6. Discussion

Air–sea gas flux estimations at regional and global scales are hindered by i) the limited data availability, and ii) the uncertainties in the gas transfer velocity parameterizations, the source of these uncertainties is—to great extent—caused by the lack of understanding about the mechanisms controlling the air–sea gas exchange. In this thesis, these limitations are addressed, first, by exploring the use of micrometeorological techniques for long-term monitoring of air–sea CH₄fluxes. And second, by evaluating the effect of relevant mechanisms controlling air–sea CO2exchange in the Baltic Sea through a sensitivity analysis of the gas transfer velocity parameterization.

The EC method, and other micrometeorological techniques, have been widely used for air–sea CO2 in oceanic and coastal applications. However, these techniques have been poorly exploited to investigate air–sea CH₄fluxes. Yang et al. (2016) presented the first results of short-term (few months) CH4 fluxes from EC in a land-based coastal station using a closed-path gas analayser. A few years later, Yang et al. (2019) presented one year of air–sea CH4using EC from the same coastal station located in the south-west coast of UK. In Paper I we analyse one year of measurements of air–sea CH₄fluxes from the Öster- garnsholm station. The results by Yang et al. (2019) using a closed-path gas analyzer, and the results presented in Paper I using an open-path system are, to the best of our knowledge, the only long-term observations of air–sea CH4

fluxes using EC. In both studies the use of the EC method for CH4 flux estimations is encouraged, and the need of using high-frequency measurements to capture the temporal variability of the flux is highlighted.

In the Baltic Sea, air–sea CH₄ fluxes have been studied using bulk methods. Gülzow et al. (2013) presented one year of air–sea CH4 methane fluxes using a ship of opportunity. The resolution of the air–sea fluxes from bulk methods is not enough to capture the daily and sub-daily variability of the fluxes. However, it may be useful to capture the spatial and seasonal variabilities, and allows—to some extent—the analysis of forcing mechanisms such as temperature, wind, upwelling events, etc. The results presented by Gülzow et al. (2013), as well as, other data reported in the MEMENTO database, are in good agreement with the results presented in Paper I. Simultaneous measurement from bulk and EC methods are necessary to fully explain the air–sea exchange processes.

The sensitivity analysis in Paper II is based on existing gas transfer velocity parameterizations. The conditions under which each parameterization was developed vary among the different studies, and the effect of the individual

(28)

mechanisms evaluated also differ from the results presented here. Norman et al. (2013b) found an enhancement of the gas transfer velocity of up to 20% in the Baltic Sea when including the parameterization suggested by Rutgersson et al. (2010). Using the same parameterization, in Paper II we found that the effect of water-side convection can represent a reduction of 73% in the annual net air–sea CO2 flux for the entire Baltic Sea. However, the effect of water-side convection has a large spatial variability among the sub-basins.

Precipitation was found to cause an increase of 27% in the net uptake of the Baltic Sea. Ashton et al. (2016) found that rain can increase the global carbon uptake by 6%. At regional scale, they observed an increment of up to 50% in the ocean carbon uptake in the Pacific Ocean. Surfactant suppression caused a reduction of the net CO₂flux of 0.7% at regional scale in the Baltic Sea and between 3.8 and 18.9% in the sub-basins. Pereira et al. (2018) found similar values (0.3-6.5%) at three coastal stations, and values ranging between 2 and 24% for the Atlantic Ocean. The gas transfer velocity parameterizations used in this study were selected based on the current resources and capabilities of the FluxEngine toolbox. In Paper II we show that the effect of water- side convection, precipitation and surfactants is a major source of the spatio- temporal variability of the fluxes in the Baltic Sea. However, other processes not considered in this study may also be relevant as controlling mechanisms of air–sea CO2exchange (i.e. wave field, ice edges, etc.).

In the global context, the mechanisms modulating the spatio-temporal variability of the air–sea CO2 fluxes are often not accounted for. Some studies have presented global carbon budgets for the marginal seas and coastal regions (Borges et al., 2005; Cai et al., 2006; Chen et al., 2013). In these estimates, the latitudinal variability is resolved, but not the seasonal and inter-annual variability.

(29)

7. Concluding remarks

Carbon dioxide (CO2) and methane (CH4) are the two most important antho- pogenic greenhouse gases. In order to understand the current climate and to be able to predict future climate scenarios in response to the increasing concentrations of greenhouse gases, it is necessary to understand the different pieces of the carbon cycle. This includes an accurate assessment of the role of the oceans on the CO2and CH4budgets.

In this work, we study the mechanisms controlling the air–sea gas exchange.

In Paper I we present one year of semi-continuous measurements of air–sea CH4fluxes. Based on the results, we suggest that the EC method is a valuable tool to evaluate and monitor CH₄fluxes in the marine environment. This technique has been very seldom used for air–sea CH4 flux estimations, and here we present the first results of this type for the Baltic Sea. The high-frequency measurements required for the flux calculations using EC are essential to capture the temporal variability of the air–sea gas exchange. Furthermore, by allowing autonomous and continuous measurements, the EC technique is suitable for long-term monitoring, therefore, capturing the seasonal variability of the flux. We also present a brief analysis of the seasonal cycle of the air–sea CH4 fluxes in the Baltic Sea. The results are validated using all available in- situ CH₄ measurements in the Baltic Sea region reported in the MEMENTO database. We show that the fluxes are mostly modulated by the concentration gradient between the sea and the atmosphere. In addition, wind speed was also shown to be a relevant parameter, in particular during the winter months when high wind speed events are common.

In Paper II we investigate the impact of relevant parameters—other than wind speed—on the air–sea CO2flux in the Baltic Sea and its four sub-basins.

We use six different gas transfer velocity parameterizations to calculate the CO2flux using a bulk method, and evaluate the effect of precipitation, water- side convection, and surfactant suppression on the flux. The seasonal cycle of the fluxes is mostly modulated by the wind and the concentration gradients. However, the use of the different parameterizations is relevant for the net seasonal and annual contributions. The implications of including the different parameters are also significant for the inter-annual and spatial variabilities.

Water-side convection seems to be the most relevant parameter modulating the flux, in particular during the winter months in the Gulf of Bothnia and Cen- tral Basin. The effect of water-side convection is noticeable as an increase on the upward (positive) flux. The effect of precipitation is noticeable throughout the year as an increase on both negative and positive fluxes, depending on the

(30)

season, but more significantly during the summer months enhancing the downward fluxes. The effect of surfactants occurs as a constrain to the downward flux during summer. With the results presented here, we highlight the importance of understanding and accounting for the relevant mechanisms in air–sea gas flux estimations in order to properly account for the spatio-temporal variability of the flux.

(31)

8. Sammanfattning på Svenska

Mekanismer som styr gasutbytet mellan hav och atmosfär i Östersjön Kol spelar en viktig roll i fysiskaliska och biogeokemiska processer i atmos- fären, biosfären och i haven. CO₂ och CH₄ är två av de vanligaste kol- föreningarna i atmosfären och de är också två av de viktigaste växthusgaserna.

Utbytet av CO₂ och CH₄ mellan hav och atmosfär är en grundläggande del i den globala kolcykeln. Utbytet styrs av koncentrationsgradienten mellan luften och havet och av verkningsgraden på överföringsprocessen. Bristen på kunskap om de styrande mekanismerna som påverkar utbytet av dessa väx- thusgaser är en stor källa till osäkerhet i uppskattningar av havens bidrag till kolflöden globalt sett. Att kvantifiera och förstå utbytesprocesserna mellan at- mosfär och hav är nödvändigt för att minska osäkerheten i uppskattningarna.

Ökad förståelse av dessa processer kan dessutom ge oss en mer kunskap om vårt nuvarande och framtida klimat.

De största osäkerheterna i uppskattningar av utbytet mellan hav och atmos- fär är kopplade till bristen på data och till osäkerheter i parametriseringar av överföringshastigheten. Kunskap om mekanismerna som styr effektiviteten på gasutbytet på lokal och regional skala är därför nödvändiga för att begränsa osäkerheterna. Den kunskap som erhållits från regionala studier är relevant, i ett globalt perspektiv, för att redogöra för bidragen från heterogena miljöer som bihav och kustområden i den globala kolbudgeten. Dessutom kan den kunskapen bidra till den allmänna förståelsen av utbytesprocesserna mellan hav och luft.

I den här avhandlingen undersöks de mekanismer som styr gasutbytet mellan atmosfär och hav i Östersjön. Användningen av mikrometeorologiska tekniker för långsiktig CH4-övervakning i kustmiljö undersöks. Ett år med kontinuerliga mätningar av CH₄-flöden mellan luft och hav med hjälp av ko- variansmätningar av turbulenta virvlar från kuststationen på Östergarnsholm tyder på att denna metod är ett värdefullt verktyg för att kunna uppskatta flö- dena av CH4i en marin miljö. Mätningarna möjliggör kontinuerlig övervakn- ing av utbytet vid höga frekvenshastigheter, vilket fångar den temporala variationen i flödet från under timmen upp till säsongsskala. Området utanför Gotland visade sig vara en nettokälla för CH4, med både koncentrationsgradienten och vindhastigheten som relevanta parametrar som styr flödet. Resul- taten stämmer väl överens med tidigare studier som tyder på att kustområden är starkt bidragande till de oceaniska CH4-utsläppen.

En känslighetsanalys av parametreringen av gasöverföringshastigheten ut- förs för att utvärdera effekten av de styrande mekanismerna som kontrollerar

(32)

utbytet av CO2 mellan luft och hav i Östersjön. Känslighetsanalysen visar att den spatio-temporala variationen av CO2-flöden är starkt styrd av konvektion i vattnet, nederbörd och ytaktiva ämnen. Konvektion i vattnet visar sig vara den viktigaste parametern som påverkar de uppåtgående flödena (hav till luft) under vintern, särskilt i Bottenviken och i den egentliga Östersjön. Un- der sommaren tenderar nederbörd och ytaktiva ämnen att styra de nedåtgående flödena (luft till hav). Effekterna från dessa faktorer är relevanta i de regionala kolbudgetarna, men kan också ha konsekvenser på global skala eftersom de för närvarande inte ingår i uppskattningarna.

(33)

9. Funding

This work was performed within the project BONUS INTEGRAL, which re- ceives funding from BONUS (Art 185), funded jointly by the EU, the German Federal Ministry of Education and Research, the Swedish Research Council Formas, the Academy of Finland, the Polish National Centre for Research and Development, and the Estonian Research Council.

The ICOS station Östergarnsholm is funded by the Swedish Research Coun- cil (grants 2012-03902 and 2013-02044) and Uppsala University.

(34)

10. Acknowledgements

I want to thank my supervisors Anna, Erik, and Marcus for your support.

Thanks for being available every time I’ve knocked on your door, for dedicat- ing me time, and effort. Thanks for the nice discussions and for sharing your knowledge with me. It feels like coming to Sweden is one of the best decisions I’ve taken, thanks for the opportunity.

Thanks to my colleagues and friends at Geo, it has been a great pleasure sharing this time with you all. You’ve been part of every step of this work;

thanks for the questions and suggestions that kept me looking for a way to improve.

I want to thank my co-authors: Hermann Bange, Annette Kock, Erik Nils- son, Jamie Shutler, Tom Holding and Gregor Rehder. I appreciate your support and co-operation; your input has been essential to this work. Special thanks to Tom, your help with FluxEngine was invaluable.

Thanks to my colleagues from the BONUS INTEGRAL project, it is a great honor to belong to such a group full of passionate and inspiring people. In particular, I want to thank those with whom I shared the joy of the INTEGRAL summer cruise, I had a great time thanks to you all.

Thanks to my family and friends, that even though we are all scattered around the world, it always feels like I have my safety net right here with me.

In particular to my parents, Lety and Marcos, and to my brother Andrés. And last but not least, I want to thank Hector ("Mi-güel") for coming all this way with me. Thank you, because only you know how to walk by my side. You are my best motivation and my best companionship.

(35)

References

Ashton, I. G., J. D. Shutler, P. E. Land, D. K. Woolf, and G. D. Quartly (2016).

“A sensitivity analysis of the impact of rain on regional and global sea–air fluxes of CO2”. In: PloS one 11.9.

Aubinet, M., T. Vesala, and D. Papale (2012). Eddy covariance: a practical guide to measurement and data analysis. Springer Science & Business Me- dia.

Baldocchi, D., M. Detto, O. Sonnentag, J. Verfaillie, Y. A. Teh, W. Silver, and N. M. Kelly (2012). “The challenges of measuring methane fluxes and concentrations over a peatland pasture”. In: Agricultural and Forest Meteo- rology153, pp. 177–187.

Baldocchi, D., E. Falge, L. Gu, R. Olson, D. Hollinger, S. Running, P. An- thoni, C. Bernhofer, K. Davis, R. Evans, et al. (2001). “FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities”. In: Bulletin of the Ameri- can Meteorological Society82.11, pp. 2415–2434.

Baldocchi, D. D., B. B. Hincks, and T. P. Meyers (1988). “Measuring biosphere- atmosphere exchanges of biologically related gases with micrometeorological methods”. In: Ecology 69.5, pp. 1331–1340.

Bange, H. W., U. Bartell, S. Rapsomanikis, and M. O. Andreae (1994). “Methane in the Baltic and North Seas and a reassessment of the marine emissions of methane”. In: Global Biogeochemical Cycles 8.4, pp. 465–480.

Borges, A. V., B. Delille, and M. Frankignoulle (2005). “Budgeting sinks and sources of CO2 in the coastal ocean: Diversity of ecosystems counts”. In:

Geophysical research letters32.14.

Cai, W.-J., M. Dai, and Y. Wang (2006). “Air-sea exchange of carbon dioxide in ocean margins: A province-based synthesis”. In: Geophysical Research Letters33.12.

Chen, C.-T., T.-H. Huang, Y.-C. Chen, Y Bai, X He, and Y Kang (2013).

“Air-sea exchanges of CO2in the world’s coastal seas”. In: Biogeosciences 10.10, p. 6509.

Dlugokencky, E. and P. Tans (2020). Trends in atmospheric carbon dioxide.

Last accessed: 08 May 2020. National Oceanic & Atmospheric Adminis- tration, Earth System Research Laboratory (NOAA/ESRL). URL:https:

//www.esrl.noaa.gov/gmd/ccgg/trends/global.html.

Esters, L., S. Landwehr, G Sutherland, T. G. Bell, K. H. Christensen, E. S.

Saltzman, S. D. Miller, and B. Ward (2017). “Parameterizing air–sea gas transfer velocity with dissipation”. In: Journal of Geophysical Research:

Oceans122.4, pp. 3041–3056.

(36)

Foken, T., M. Aubinet, and R. Leuning (2012). “The eddy covariance method”.

In: Eddy covariance. Springer, pp. 1–19.

Friedlingstein, P., M Jones, M. O’Sullivan, R Andrew, J. Hauck, G Peters, W. Peters, J. Pongratz, S. Sitch, C. Le Quéré, et al. (2019). “Global carbon budget 2019”. In: Earth System Science Data 11.4, pp. 1783–1838.

Gattuso, J.-P., M. Frankignoulle, and R. Wollast (1998). “Carbon and car- bonate metabolism in coastal aquatic ecosystems”. In: Annual Reviews of Ecolology and Systematics29, pp. 405–434.

Gülzow, W, G. Rehder, J. Schneider von Deimling, S Seifert, and Z. Tóth (2013). “One year of continuous measurements constraining methane emissions from the Baltic Sea to the atmosphere using a ship of opportunity”.

In: Biogeosciences (BG) 10, pp. 81–99.

Ho, D. T., C. S. Law, M. J. Smith, P. Schlosser, M. Harvey, and P. Hill (2006).

“Measurements of air–sea gas exchange at high wind speeds in the South- ern Ocean: Implications for global parameterizations”. In: Geophysical Re- search Letters33.16.

Högström, U., E. Sahlée, W. M. Drennan, K. K. Kahma, A.-S. Smedman, C. Johansson, H. Pettersson, A. Rutgersson, L. Tuomi, F. Zhang, and M.

Johansson (2008). “Momentum fluxes and wind gradients in the marine boundary layer: A multi platform study”. In: Boreal environment research 13.6, pp. 475–502.

Holding, T., I. G. Ashton, J. D. Shutler, P. E. Land, P. D. Nightingale, A. P.

Rees, I. Brown, J.-F. Piolle, A. Kock, H. W. Bange, et al. (2019). “The FluxEngine air–sea gas flux toolbox: simplified interface and extensions for in situ analyses and multiple sparingly soluble gases”. In: European Geosciences Union (EGU).

IPCC (2013). Climate change 2013: The physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press.

Jähne, B, W Huber, A Dutzi, T Wais, and J Ilmberger (1984). “Wind/wave- tunnel experiment on the Schmidt number—And wave field dependence of air/water gas exchange”. In: Gas transfer at water surfaces. Springer, pp. 303–309.

Jahnke, R. A. (2010). “Global synthesis”. In: Carbon and nutrient fluxes in continental margins. Ed. by K.-K. Liu, L. Atkinson, R. Quiñones, and L.

Talahue-McManus. Berlin Heidelberg: Springer, pp. 597–616.

Kaimal, J. C. and J. J. Finnigan (1994). Atmospheric boundary layer flows:

their structure and measurement. Oxford university press.

Kock, A. and H. W. Bange (2015). “Counting the ocean’s greenhouse gas emissions”. In: Eos 96.3, pp. 10–13.

Liss, P. S. and P. G. Slater (1974). “Flux of gases across the air–sea interface”.

In: Nature 247, pp. 181–184.

Mechanisms controlling air-sea gas exchange in the Baltic Sea