Variability of summer CH4 and CO2 flux rates in and between three large Swedish lakes : A spatio-temporal field study

Department of Thematic Studies

Campus Norrköping

Bachelor of Science Thesis, Environmental Science Programme, 2020

Nilsson, Isak

Seifarth, Filip

Variability of summer CH

4

and

CO

2

flux rates in and between

three large Swedish lakes

A spatio-temporal field study

Rapporttyp Report category Licentiatavhandling Examensarbete AB-uppsats C-uppsats D-uppsats Övrig rapport Språk Language Svenska/Swedish Engelska/English Titel

Sommarvariation i CH4- och CO2-flöden i och mellan tre stora svenska sjöar: En rumslig och tidsmässig fältstudie Title

Variability of summer CH4 and CO2 flux rates in and between three large Swedish lakes: A spatio-temporal field study Författare

Authors

Nilsson, Isak Seifarth, Filip

Sammanfattning

Förståelsen av naturliga växthuscykler är avgörande för att göra budgetar för växthusgaser, eftersom dessa uppskattningar och budgetar agerar som grund i policyprogram för arbete med klimatförändringar och global uppvärmning. Sjöar har endast nyligen inkluderats i globala växthusgodsbudgetar som en källa för växthusgasutsläpp, och de flesta studier av flödeshastigheter genomförs på sjöar <10 km2_{, vilka endast utgör}

ungefär hälften av den globala sjöarealen – vilket leder till att data om växthusgasflöden från stora sjöar saknas. CO2 och CH4 är de mest potenta

växthusgaserna, och sjöar hyser produktionsprocesser samt tar emot dessa gaser från kringliggande miljöer. Denna studie nyttjade en kammarmetod med CO2-sensorer för att studera CH4 och CO2-flödeshastigheter från tre stora svenska sjöar. Detta gjordes dels genom att ankra kammare på grunt

djup, och dels genom att låta kammare driva passivt på öppet vatten. Provtagningen genomfördes under två perioder sommaren 2019, slutet av juni-början av juli och augusti. För CH4 hittades rumslig skillnad mellan djupa och grunda transekter i sjöar, men ingen tidsmässig skillnad hittades

mellan studieperioder. Skillnader mellan sjöar i de djupa och grunda kammargrupperna hittades. Ett möjligt fall av metanbubblor från djupt vatten registrerades, liksom en korrelation mellan CH4-flödeshastighet och vattentemperatur. För CO2 hittades ingen skillnad mellan djupa och grunda

kammare eller mätperioder. En skillnad i den djupa kammargruppen hittades mellan två av sjöarna, trots att alla tre var av olika storlek, djup och trofiklass. Studiens resultat indikerar att koncentrationer av CO2 i vatten var nära mättnad med atmosfären under studieperioderna. Ingen korrelation

mellan CO2-flödeshastighet och vattentemperatur observerades. Oväntade småskaliga variabilitetsmönster i CO2-flöde observerades medan kamrar

drev passivt. Medan vissa observerade mönster för de två gaserna kan förklaras av tidigare kunskap, kan andra av våra observationer inte förklaras av tidigare litteratur, och detta tydliggör behovet av fortsatt forskning på växthusgasflöden från stora sjöar.

Abstract

Understanding of natural greenhouse gas (GHG) cycles is crucial for making GHG budgets, which work as basis in climate change and global warming policy programs. Lakes as a source for GHG activity have only recently been included in global GHG budgets, and most studies of lake GHG flux rates are conducted on lakes <10 km2_{, which only comprise roughly half of the global lake area—making data of large lake flux rates}

scarce. CO2 and CH4 are the primary contributors of GHGs, and lakes house production processes and receive these gasses via lateral transport. This

study utilized a floating chamber method with CO2 sensors to study CH4 and CO2 flux rates from three large Swedish lakes. To do this, chambers

were anchored at shallow depth, as well as passively drifted on open water. Sampling was conducted during two periods in the summer 2019, late June–early July and August. For CH4, spatial difference was found between deep and shallow transects within lakes, no temporal difference was

found between study periods. Difference between lakes within the deep and shallow chamber groups was found. One possible instance of deep-water ebullition was recorded, and a correlation between CH4 flux rate and water temperature was observed. For CO2, no difference between deep and

shallow chambers or measurement periods was found. One instance within the deeper chamber group was found to be different between two of the lakes, despite all three lakes being of different size, depth and trophic state. The study results indicate CO2 water concentrations near saturation with

atmosphere during the sampling periods. No correlation between CO2 flux rate and water temperature was observed. Unexpected small-scale

variability patterns in CO2 flux was observed while chambers were passively drifting. While some observed patterns for the two gases could be

explained by previous findings, some of our observations could not be explained on the basis of previous literature, highlighting the need for further study of GHG flux rates from large lakes.

ISBN _____________________________________________________ ISRN LIU-TEMA/MV-C—20/25—SE _________________________________________________________________ ISSN _________________________________________________________________ Handledare Tutor David Bastviken Examinator Examiner Sofie Storbjörk Nyckelord

Stora sjöar, sötvatten, flödeskammare, utsläpp, bubblor, diffusion, kolcykel, kolbudget, metanbudget, växthusgaser, metan, CH4, koldioxid, CO2,

Datum

Date 2020-06-12

URL för elektronisk version

http://www.ep.liu.se/index.sv.html

Institution, Avdelning

Department, Division Tema – Miljöförändring, Miljövetarprogrammet

Department of Thematic Studies – Environmental change, Environmental Science Programme

Abstract

Understanding of natural greenhouse gas (GHG) cycles is crucial for making GHG budgets, which work as basis in climate change and global warming policy programs. Lakes as a source for GHG activity have only recently been included in global GHG budgets, and most studies of lake GHG flux rates are conducted on lakes <10 km2, which only comprise roughly half of the global lake area—making data of

large lake flux rates scarce. CO2 and CH4 are the primary contributors of GHGs, and lakes house

production processes and receive these gasses via lateral transport. This study utilized a floating chamber method with CO2 sensors to study CH4 and CO2 flux rates from three large Swedish lakes. To do this,

chambers were anchored at shallow depth, as well as passively drifted on open water. Sampling was conducted during two periods in the summer 2019, late June–early July and August. For CH4, spatial

difference was found between deep and shallow transects within lakes, no temporal difference was found between study periods. Difference between lakes within the deep and shallow chamber groups was found. One possible instance of deep-water ebullition was recorded, and a correlation between CH4 flux

rate and water temperature was observed. For CO2, no difference between deep and shallow chambers

or measurement periods was found. One instance within the deeper chamber group was found to be different between two of the lakes, despite all three lakes being of different size, depth and trophic state. The study results indicate CO2 water concentrations near saturation with atmosphere during the

sampling periods. No correlation between CO2 flux rate and water temperature was observed.

Unexpected small-scale variability patterns in CO2 flux was observed while chambers were passively

drifting. While some observed patterns for the two gases could be explained by previous findings, some of our observations could not be explained on the basis of previous literature, highlighting the need for further study of GHG flux rates from large lakes.

Sammanfattning

Förståelsen av naturliga växthuscykler är avgörande för att göra budgetar för växthusgaser, eftersom dessa uppskattningar och budgetar agerar som grund i policyprogram för arbete med klimatförändringar och global uppvärmning. Sjöar har endast nyligen inkluderats i globala växthusgodsbudgetar som en källa för växthusgasutsläpp, och de flesta studier av flödeshastigheter genomförs på sjöar <10 km2, vilka

endast utgör ungefär hälften av den globala sjöarealen – vilket leder till att data om växthusgasflöden från stora sjöar saknas. CO2 och CH4 är de mest potenta växthusgaserna, och sjöar hyser

produktionsprocesser samt tar emot dessa gaser från kringliggande miljöer. Denna studie nyttjade en kammarmetod med CO2-sensorer för att studera CH4 och CO2-flödeshastigheter från tre stora svenska

sjöar. Detta gjordes dels genom att ankra kammare på grunt djup, och dels genom att låta kammare driva passivt på öppet vatten. Provtagningen genomfördes under två perioder sommaren 2019, slutet av juni-början av juli och augusti. För CH4 hittades rumslig skillnad mellan djupa och grunda transekter i sjöar,

men ingen tidsmässig skillnad hittades mellan studieperioder. Skillnader mellan sjöar i de djupa och grunda kammargrupperna hittades. Ett möjligt fall av metanbubblor från djupt vatten registrerades, liksom en korrelation mellan CH4-flödeshastighet och vattentemperatur. För CO2 hittades ingen skillnad

mellan djupa och grunda kammare eller mätperioder. En skillnad i den djupa kammargruppen hittades mellan två av sjöarna, trots att alla tre var av olika storlek, djup och trofiklass. Studiens resultat indikerar att koncentrationer av CO2 i vatten var nära mättnad med atmosfären under studieperioderna. Ingen

korrelation mellan CO2-flödeshastighet och vattentemperatur observerades. Oväntade småskaliga

variabilitetsmönster i CO2-flöde observerades medan kamrar drev passivt. Medan vissa observerade

mönster för de två gaserna kan förklaras av tidigare kunskap, kan andra av våra observationer inte förklaras av tidigare litteratur, och detta tydliggör behovet av fortsatt forskning på växthusgasflöden från stora sjöar.

Preface

Before the reader indulges in this bachelor thesis, we would like to express our thanks and gratitude to those who made this possible. First and foremost, thank you David Bastviken for sharing your knowledge, passion and concern about the subject. Thank you for being a good supervisor and teacher throughout the project and our years at Linköping University. A big thanks to Ingrid Sundgren for performing the laboratory analysis, for being supportive and patient during the field campaign, as well as always answering our questions. Thank you to Jonathan Schenk for helping us with the stubborn sonar log files. We would also like to thank those closest to us for supporting and assisting in finalizing the thesis.

This study has truly been a learning experience and personally developing for both of us, and we now proudly present the final product.

Keywords

Large lakes, Freshwater, Flux chambers, Emissions, Ebullition, Diffusion, Carbon cycle, Carbon budget, Methane budget, Greenhouse gases, GHGs, Methane, CH4, Carbon dioxide, CO2, Variability, Spatial,

Table of contents

1. Introduction

3

1.1 Aim and research questions

4

1.2 Delimitations

4

2. Background

5

2.1 Gas properties

5

2.1.1 Methane 5

2.1.2 Carbon dioxide 6

2.2 Flux to the atmosphere

8

2.2.1 Methane 8

2.2.2 Carbon dioxide 10

3. Materials and methods

11

3.1 Method and setup

11

3.1.1 Measuring period and times 11 3.1.2 Chamber transects and layout 12 3.1.3 Sampling method and vial transfer 13 3.1.4 CO2 Sensor method 14

3.2 Study sites

14

3.3 Analysis method

18

3.4 Data treatment and flux calculations

18

3.4.1 Methane data 18 3.4.2 Carbon dioxide data 19

3.4.3 Other data 20

3.5 Statistics

20

4. Results

21

4.1 Methane

21

4.1.1 Within lake variation 21 4.1.2 Between-lake variation 22 4.1.3 Flux correlation with environmental variables 23

4.2 Carbon dioxide

24

4.2.1 Within lake variations 24 4.2.2 Between-lake variation 25 4.2.3 Flux correlation with environmental variables 26

5. Discussion

28

5.1 Method evaluation

28

5.2 Methane

30

5.2.1 Spatiotemporal variations 30 5.2.2 Correlation with environmental variables 32 5.2.3 Methane flux compared with previous studies 32

5.3 Carbon dioxide

33

5.3.1 Spatiotemporal variations 33 5.3.2 Correlation with environmental variables 34 5.3.3 Carbon dioxide flux compared with previous studies 34

6. Conclusions

36

1. Introduction

An unavoidable challenge facing the world’s population is climate change, and the main driver for climate change is global warming. Several gases have been found to have a radiative effect—their presence in the atmosphere traps heat waves that would otherwise leave the planet—hence the term greenhouse gases (GHGs) (Field, Raupach and Victoria, 2004). The two primary GHGs contributing to global warming are carbon dioxide (CO2) and methane(CH4). The atmospheric CO2 concentration was

stable, varying with <20 ppm, for the past 11 000 years until the industrial revolution (IPCC, 2013). Since the 1950s temperature changes in the ocean and atmosphere have been observed, correlating with higher concentrations of greenhouse gases in the atmosphere (IPCC, 2013). Linkage between the warming effects of GHGs and the usage of fossil fuels has led to extensive research regarding the issue of global warming (Field, Raupach and Victoria, 2004).

When studying global warming, a useful tool is GHG budgeting. GHG budgets are based on the knowledge and analysis of physical feedbacks in models of natural biogeochemical cycles. GHG budgets can provide estimations of global warming results from different sets of anthropogenic GHG emission scenarios, explore climate sensitivity, and they can be used to determine GHG emission limits that align with global warming targets. Past work on making GHG budgets has clarified the need to quantify and assess the climate sensitivity, not only of anthropogenic GHG emissions but also on natural background fluxes, to allow reliable predictions of the future climate (IPCC, 2013).

Terrestrial wetlands and marshes have been studied and considered key environments since the early days of biogeochemical studies (Bartlett and Hariss, 1993). However, environments that have previously not been studied or considered important are lakes (Bastviken et al., 2004; Cole et al., 2007). Lakes are also integrated in the terrestrial environment and with an increasing amount of studies on lakes, it is clear that these biogeochemically intense ecosystems (Verpoorter et al., 2014) are important for the terrestrialGHG cycling (Casper et al., 2000; Emmerton et al., 2016; Erkkilä et al., 2018). Freshwater lakes cover only 3.7% of terrestrial area (Verpoorter et al., 2014) but are estimated to emit GHGs disproportionally compared to the other terrestrial environments (Bastviken et al., 2011; Natchimuthu, 2016). Lakes house processes that produce—and laterally receive—both CH4 and CO2, and are thus

usually supersaturated with these gasses which ultimately are released to the atmosphere (Natchimuthu, 2016). Accurate estimates of these flux rates in lakes are important for improving global GHG budgets and making estimates for prospective global warming (DelSontro, Beaulieu and Downing, 2018). Many of the lake GHG flux studies have been conducted on small lakes, making data for large lakes scarce, at least for CH4. This means that in global estimates, the data collected on small lakes are

extrapolated to represent all lakes, including large lakes (Bastviken et al., 2004, 2011). Advances in remote sensing have enabled better understanding of accurate lake areas and occurrence of different lake sizes. (Verpoorter et al., 2014) estimates a total global lake area of all lakes > 2000 m2 of roughly 5*106

km2. Larger lakes, which (Verpoorter et al., 2014) defines as lakes with an area >10 km2, are fewer in

number, which is one reason for these previously being neglected. However, their area totals over 2*106

km2, which is close to half of the total global lake area (Verpoorter et al., 2014). This means that in

estimates, the extrapolated data is relevant for only half of the global lake area.

Hence, there are reasons to study CH4 and CO2 flux rates from large lakes to contribute to the

understanding of how these systems behave in terms of GHG sinks or emitters. This would provide more accurate data where there were previously extrapolations, improving global GHG budget estimates and ultimately supporting better climate change and global warming policy programs.

1.1 Aim and research questions

In light of the knowledge gap and data deficiency the aim of this thesis is to study fluxes of CH4 and

CO2 to and from three large Swedish lakes focusing on spatial and temporal variabilities, to contribute

to the understanding of the variations in large lake gas flux rates. The key questions are:

– How does CH4 and CO2 flux rates vary spatially and temporally, within and between different

lakes?

– Can flux rate variability be correlated with water temperature?

1.2 Delimitations

The primary delimitation is studying only temperature as an environmental variable affecting flux rates. This determination was made, despite other variables being accessible, mainly for reasons of data quality and available resources within the framework of a bachelor thesis. Depth is clearly an interesting variable that could affect flux rates. However, our depth data is not suitable for studying correlation with flux. The way in which depth data is collected, at deployment and collection of flux chambers, it does not contain any information on chamber depth between those moments. As chambers were free to move with the wind, they could move during the deployment period. Furthermore, the chambers were mounted across a cord spanning roughly 15 m between the outermost chambers, and even within this distance, the depth could vary greatly. Similarly, as the air pressure data used in this study is not primary data, it is considered unsuitable for analysis of correlation with flux rate. As discussed in the background, lateral transport is important for both GHGs, studying correlations with this is something that we could not fit within the timeframe of a bachelor thesis.

The temporal variability in this study is delimited to the study days during the two study periods. These are not necessarily representative for whole months or seasons. They are established to study variability in time, not differences between months or seasons—for which nothing can be said based on our data.

2. Background

2.1 Gas properties

2.1.1 Methane

CH4 is a radiative trace gas, and one of the main products from decomposition of organic matter (OM)

in freshwater environments. It is an odorless gas that on a molecular level has a greenhouse warming potential 28 times that of CO2, based on radiative properties and residence time, on a 100-year timescale.

Its in-atmosphere lifetime is estimated between eight to twelve years. Most of the CH4 in the atmosphere

has biogenic origin and it accounts for roughly 20% of increased greenhouse effect observed the last ~300 years (IPCC, 2013). CH4 is relatively insoluble, the molar solubility in freshwater is 1.6 mol m-3

at 20 °C (molar solubility is how many moles that can be dissolved per volume unit before a solution becomes saturated) (Casper et al., 2000).

The biological production process of CH4 is called methanogenesis. Methanogenesis is performed by

microbial archaea called methanogens (Conrad, 1996). Archaea is its own branch of microbes, like bacteria and eukaryotes (Pace, 2006). There are two principal types of methanogenesis, one that is based on acetate, and one that is based on hydrogen. In the former case, acetate (CH3CO2) is split to CH4 and

CO2. In the latter case hydrogen (H2) reacts with CO2 to form CH4 and H2O. These two processes can

often happen in parallel (Bastviken, 2009). There are other types of methanogenesis, but the acetate- and hydrogen-based ones are believed to be the primary contributors (>95%) to CH4 production in

sediments (Segers, 1998).

Methanogenesis is a terminal anoxic (water with little to no dissolved oxygen) degradation step of OM— it is a last degradation step—(see Figure 1) and for it to be possible, more complex OM has to be degraded to a level where it is available to the methanogens (Bastviken, 2009). There are other terminal anoxic degradation steps for OM, which utilize other oxidized compounds that can act as electron acceptors in chemical reactions. These often yield more energy than methanogenesis, which ultimately means that in presence of other electron acceptors methanogenesis is inhibited (Segers, 1998; Bodegom and Stams, 1999; Conrad, 2002).

Figure 1. Steps of anaerobic (a process in the absence of oxygen) degradation of organic matter (adapted

Methanogenesis is affected by other factors in the surrounding environment as well. For example, a 10 °C temperature increase results in about four times higher methanogenesis (Segers, 1998; Conrad, 2002; Whalen, 2005). Available newly produced OM consequently has a net positive effect on methanogenesis, since it feeds the methanogens (Segers, 1998; Whalen, 2005). Roots from plants have a double effect, positive from leakage of OM and decay of the roots, but a negative effect from leakage of oxygen (O2) (Segers, 1998; Conrad, 2002). The net effect of roots on methanogenesis is however

presumed to be positive (Bastviken, 2009). Previously methanogens were thought to be absolute anaerobic, but it is now known that they tolerate considerable exposure to O2, though it inhibits the

methanogenesis process (Conrad, 2002). There is currently an ongoing debate regarding aerobic (a process where oxygen is required) CH4 production in lakes (Tang et al., 2016; Khatun et al., 2019),

however this is not yet established knowledge.

Oxidation of CH4 can occur via both aerobic and anaerobic processes, however in freshwater

environments aerobic oxidation generally dominates. This process is performed by methane-oxidizing bacteria (MOB). It takes place at the interface between the oxic (water containing dissolved oxygen) and anoxic environments—where CH4 is available in the anoxic—and the oxidation takes place in the

oxic since it is an aerobic process. This border area can exist in the top layers of the sediment, deeper in the sediment close to roots leaking O2 or in the water itself (Bastviken, 2009).

CH4 concentrations in water are a function of the methanogenesis, the oxidation, and flux rates to and

from the water. Concentrations are also affected by water turbulence and mixing, well-mixed oxic waters can have concentrations a hundred times lower than anoxic parts in stratified (layered) water (Bastviken, 2009).

2.1.2 Carbon dioxide

CO2 is the most common GHG and for plants it is crucial for photosynthesis. CO2 is more soluble than

CH4 with a molar solubility of 39 mol m-3 at 20 °C in freshwater. This leads to the accumulation of

higher concentrations of CO2 than CH4 at greater depths (Casper et al., 2000). CO2 concentration in

freshwater lakes is often supersaturated in relation to the atmosphere. This surplus of CO2 is often

accredited to allochthonous (originated from external sources) organic carbon (OC) from land sources (Dillon and Molot, 1997; Huotari et al., 2011).

Allochthonous OC is transported to lakes via surrounding soil and lateral input (Tranvik et al., 2009) and fuels much of the OC mineralization that takes place in lakes (see Figure 2). OC mineralization is the process in which heterotrophic organisms transforms OC into inorganic carbon (IC), CO2 being a

part of this IC (Gudasz et al., 2010; Cardoso et al., 2019). This OC mineralization often takes place by the sediment surface and then OC that does not get mineralized instead gets buried in the sediment. OC sedimentation is dependent on oxygen exposure (Gudasz et al., 2010) and OC mineralization is dependent on available electron acceptors, temperature, pH, mixing regime of the water column and IC concentration (Cardoso et al., 2019).

Varied results from studies place sedimentary C contribution to total release of CO2 from the lake

anywhere from high to low, but is generally believed to be an important contributor (Chmiel et al., 2016). Once again, this depends on different characteristics, for example temperature (higher temperatures yield elevated OC mineralization) and mineralization rate in anoxic environments are lower than in oxygenated environments (Chmiel et al., 2016).

Lakes receive allochthonous IC from surrounding soil, atmosphere and lateral inputs (Tranvik et al., 2009; Natchimuthu, 2016). IC is made up of bicarbonate ions, carbonate ions and CO2 (Cole and Prairie,

2009). When CO2 is in the lake, most of it is emitted to the atmosphere via diffusive flux, laterally

transported or gets buried in the sediment (Tranvik et al., 2009). IC transport is dependent on lake characteristics like nutrient conditions, lake morphology (size, volume and shape of a lake) and regional climate. In boreal lakes (like the ones in this study), lateral transport of CO2 to the lake is sometimes

highly significant (Engel et al., 2018).

There is some autonomous (originated from the lake itself, not from external sources) CO2 produced

within a lake. This autonomous CO2 stems from respiration of heterotrophic organisms. While studies

have found some lakes where primary production is dominant, most studies have found respiration to dominate over primary production (Sobek, 2005), further fueling CO2 emissions from lakes. CO2 can

also chemically react when dissolved in water, in a process called hydration where CO2 and hydroxide

(OH-) react and create bicarbonate (Bade, 2009).

Figure 2. Pathways for OC, IC and CO2 in a lake. Depending on if the water is supersaturated or

undersaturated, lakes emit or receive CO2 from the atmosphere.

Depending on the number of streams and how turbulent the water is, advective transport of C can occur. A study made by Tranvik et al. (2009) showed that out of six lakes studied, four of them transported over 65% of C downstream, clearly making this an important factor for how much C that stays in the lake. Shorter water residence in a lake means that less C stays in the lake, limiting OC mineralization (Engel et al., 2018) and CO2 flux to atmosphere or burial to sediment.

According to Verpoorter et al. (2014), out of 117 million lakes, 90 million of these are in the smaller group of lakes with an area of 0.002 to 0.01 km2. These lakes are shallow and plenty of light reaches the

lakes, thus making them some of the most productive systems on earth (Tranvik et al., 2009). Larger lakes have less data recorded (Bastviken et al., 2004) and since they are different from smaller lakes, OC transport should behave differently from smaller lakes. This should in turn affect CO2 flux rates,

2.2 Flux to the atmosphere

2.2.1 Methane

At least four different pathways contribute to CH4 exchange between water and the atmosphere. Three

of these pathways include aquatic plants transporting sedimentary CH4 to the atmosphere, diffusive flux

at the water-atmosphere interface, and CH4 bubbles rising to the surface (ebullition) due to gas

accumulation resulting in a buoyancy allowing the bubble to break free from its place in the sediment (Casper et al., 2000; Bastviken, 2009). The fourth pathway for CH4 flux to atmosphere is from diffusive

flux during certain circumstances, where the amount of diffusive flux increases by several magnitudes. This happens when stratified water completely mixes (lake overturn), CH4 stored in the anoxic segment

of the stratified water enters the oxic segment of water and the air-water interface. When this happens the water CH4 concentration becomes several times higher than usual. Since this diffusive flux is faster

due to the higher concentrations, less of the CH4 is oxidized than at non-overturn diffusive flux (Fallon

et al., 1980; Bastviken, 2009). The four pathways are illustrated in Figure 3.

Figure 3. CH4 flux pathways in stratified freshwater lake (adapted from Bastviken, 2009).

In many aquatic plants there are gas transport channels to transport O2 to the roots, and in these channels,

CH4 can travel the opposite way—to the atmosphere. Thus, the plants act as chimneys, letting the CH4

bypass oxic environments and oxidation, allowing quick transport directly to atmosphere. For some plants there are daily patterns in CH4 transport due to sunlight and heating. Observations also show less

ebullition in areas with aquatic plants that grow past the water surface (Bastviken, 2009). However, the plant mediated CH4 flux is not further addressed in this thesis, since it only studies the open water flux

pathways.

Diffusive flux is driven by turbulent diffusion transport of CH4 molecules from the sediment to the

surface through the water column. Since this process is slow, large parts of the CH4 transported from

the anoxic sediment trough oxic parts are oxidized by the aerobic MOB (Bastviken, 2009).

Turbulent diffusion, that drives diffusive CH4 flux, is commonly modelled as fickian transport, and can

in one dimension be quantified by the equation,

𝐽 = −𝐷 (𝑑𝐶_𝑑𝑋), (Eq 1) where J is flux density in M/L2T, D is the fickian mass transport coefficient in L2/T and d is momentary

change, difference. C is CH4 concentration in M/L3, X is distance for which concentration change is

When CH4 has reached the surface water by turbulent diffusion transport, the exchange with the

atmosphere is driven by the difference in concentration. The exchange can be calculated with Fick’s law of diffusion, which describes the dependency of gas exchange on physical and chemical properties. The physical and chemical parameters of the lake affect the gas exchange, but it can be simplified with the equation,

F =k(Cs u r−Ce q), (Eq 2)

where F is the flux and k is the piston velocity (Cole et al., 2010). The piston velocity can be described as the rate equilibrium is reached of the concentration gradient (Bade, 2009). Csur and Ceq are

concentration at the water surface and the air concentration respectively. Knowing temperature and air pressure, these can easily be obtained from measurements while k is more complicated. This in theory is a function of turbulent energy exchange between atmosphere and surface water but is difficult to assess in reality (Cole et al., 2010).

Ebullition is, as mentioned above, the release of bubbles from the sediment. When enough CH4 has

accumulated to overcome the resistance for bubble release, or the resistance decreases, the bubble can break free from the sediment and rapidly rise through the water pillar and enter the atmosphere. Since ascent is rapid, most CH4 is not oxidized (Bastviken, 2009). In cases of ebullition from deep waters,

parts of the bubble can dissolve in water during ascent to later oxidize (McGinnis et al., 2006).

The three different open water flux types correlate with lake size, and some vary more than others, as shown in Figure 4. Ebullition is thought to always be a significant contributor to the total CH4 flux from

open water pathways and does not vary drastically with lake size. Storage flux and diffusive flux at lake overturn events contributes more than continuous diffusive flux in lakes smaller than roughly 100 000 m2 (0.1 km2). Continuous diffusive flux is the larger pathway of the two in lakes larger than that, due to

the amount of actual surface area that is exposed to the atmosphere and allows for diffusive flux (Bastviken, 2009).

Figure 4. Relative open water CH4 flux pathways against lake size (adapted from Bastviken, 2009).

As shown in Equation 2 above, diffusive flux is a function of the difference in concentration and the gas piston velocity in the water. These are consequently the factors that affect the diffusion rates; water CH4

concentration, water movement close to the surface and wind that dilute the concentration in the air segment immediate to the water surface (Bastviken et al., 2004). Since the available CH4 ultimately

contributes to the CH4 concentrations in the water, water temperature could also indirectly affect

Since ebullition is affected by pressure, it is also depth dependent. Ebullition is consequently more common in shallow depth where the hydrostatic pressure is low. However, it can also occur at greater depths at locations with steep lake bottoms making OM rapidly accumulate and degrade—creating hotspots for methanogenesis and consequently more intense ebullition (Sobek et al., 2012). The pressure dependency also means weather, namely the above air pressure, affects ebullition. Observations show correlation between ebullition episodes and frontal air pressure drops (Bastviken, 2009). Other events can trigger ebullition as well, disturbances in the sediment that reduce the resistance for bubble release such as waves causing turbulence close to the sediment was found correlating with ebullition by Joyce and Jewell (2003). Temperature does, as mentioned above, affect methanogenesis. Therefore, it can also affect ebullition (Duc, Crill and Bastviken, 2010).

2.2.2 Carbon dioxide

The majority of CO2 emission from freshwater lakes is diffusive and usually almost no CO2 is ebullitive

(Casper et al., 2000; Natchimuthu, 2016). CO2 has higher water solubility than CH4 and thus rarely form

bubbles. When the lake is supersaturated with CO2 the concentration gradient between the lake and air

becomes greater. The concentration gradient along with the gas exchange velocity can be used to calculate the transport across water-air (Martinsen, Kragh and Sand-Jensen, 2020). This can be calculated with Fick’s law of diffusion (see Equation 2).

The general parameters affecting CH4 exchange at the water surface applies for CO2 as well, i.e., section

2.2.1. These parameters are many and complex (Cole et al., 2010) making more specific measurements required for individual lakes to make estimates more accurate. For example, organic compounds can sometimes occur in surface slicks, reducing the impact of wind which stirs up the water (Bade, 2009).

3. Materials and methods

3.1 Method and setup

To measure CH4 and CO2 fluxes, a floating chamber method was used. The method has been verified to

not bias flux measurements (Cole et al., 2010; Gålfalk et al., 2013) and has become popular for measuring GHG flux from open water (see Bastviken et al., 2004, 2010, 2011; Cole et al., 2010; Gålfalk

et al., 2013; Podgrajsek et al., 2014; West, Creamer and Jones, 2016; Erkkilä et al., 2018; Paranaíba et al., 2018; Hallgren and Åman, 2019). In this study, the method was applied in an alternate way—

chambers were allowed to passively drift with the wind on the open water in the middle of the lake. This was done for several reasons, primarily to allow study of flux behavior at open large lake areas which include effects of water movements, wind speeds at open waters, greater depths, levels of organic matter and other dynamics. It was also done for practical reasons, anchoring chambers was not possible in all lakes, therefore it was important to conduct all deeper water measurements the same way on all lakes. The method specifics differed slightly for the two different gasses. For CH4, a conventional flux chamber

headspace measurement as presented in Bastviken et al. (2004) was used. For CO2, sensors inside flux

chambers were used (described further in segment 3.1.4). Chamber details are developed in segment 3.1.2. Unlike Bastviken et al. (2004, 2010) the measurements were only performed for short periods of time. Measurement length does not bias the results, however a 50% higher flux rate during daytime has been observed (Bastviken et al., 2004). Measurements were only conducted during daytime in this study to attend chambers at all times, due to frequent chamber leakage and corresponding data loss on large lakes when chambers were deployed unattended overnight (Beijer and Skoglund, 2019). On large lakes the weather can vary greatly during a day, and the chamber measurements were easily disrupted by waves, causing either accidental ventilation of chambers or even completely flipped chambers. Accidental ventilation resulted in unusable data and flipped chambers in addition caused sensor malfunction. This led to the decision to only sample during the day while field personnel could monitor chambers, and abort sampling early in case of weather getting worse, to secure the data already gathered.

3.1.1 Measuring period and times

Initially, the aim was to sample all three lakes three times during each period, since that would require nine field days, and each period was at least two weeks long. The weather quickly obstructed those plans, forcing the decision to lower the ambitions. Two sampling days on each lake became the new goal. To not create a skewness in data, sampling each lake the same amount of days was preferable. During the second sampling period, the weather prevented sampling on Vättern a second day. In Table 1 below, all dates when data was collected are listed, divided in the two measurement periods. The lakes are further described in section 3.2 below.

Table 1. Table showing which lake sampled at what date in the two measurement periods during the summer

of 2019.

Period 1

Period 2

Date Lake Date Lake June 24th Roxen August 5th Roxen

June 27th Glan August 6th Glan

June 28th Vättern August 8th Glan

July 6th Glan August 9th Vättern

July 10th Roxen August 15th Roxen

3.1.2 Chamber transects and layout

Initially, three transects of five chambers were used. Two for placing at shallow depth to catch the hard-to-measure ebullition, and one for drifting at greater depths. Specific chamber information such as area, volume and other construction details can be found in the supporting information to Bastviken et al. (2015). After the first day the decision to exclude one of the transects was made. This decision was taken due to the unreasonable amount of chamber management and sample collection time needed. Previous studies have been working with even greater number of chambers, however usually on smaller and calmer lakes making chamber handling much easier than on these large lakes. When discussing the purpose of the study, it was deemed suitable to extend the drifting transect with two more chambers to collect as much data as possible while keeping chamber management on a reasonable level. This extension of the drifting transect was first implemented on June 6th, from that point the study used five

transect chambers and seven drifting chambers at all times. Two chambers in each transect was fitted with a CO2 sensor. This resulted in it usually being the following representing data for each sampling

day: for CH4 each chamber was one representation of flux, resulting in five representations of shallow

flux, and seven representations of deeper CH4 flux. For CO2 the sensors were the representations of flux,

resulting in two representations of shallow CO2 flux and two representations of deeper fluxes.

Transect chambers

The chambers that were fixed at shallow depth had the purpose of measuring shallow depth/close to shore fluxes of both CH4 and CO2, as fluxes from shallow depths differ (more ebullition, among other

things) from those at greater depths (Bastviken et al., 2004). These chambers are in this thesis called “transect chambers” or just “transect”. The transect consisted of five chambers in total, with CO2-sensors

in the chambers at either end of the transect. The transect was anchored at a suitable distance from the shoreline, with margin for changing wind and the transect shifting direction with the wind and waves. The anchor was placed at a depth of approximately 1–4 m, depending on lake, that met these requirements. The chambers were then allowed to stretch out in the wave direction from the anchor placement. Between chambers the cord length was ~3.5 m, while the cord length between the last chamber and the anchor float was ~5 m. The length from anchor float to anchor varied depending on lake depth, it was ~2 m for Glan and Roxen, while being 4 m for Vättern.

Drifting chambers

The chambers that were fixed to the boat and drifted with the wind, had the primary purpose of measuring open water flux of both CH4 and CO2, and are in this thesis referred to as “drifting chambers”

or just “drift”. The drifting chambers also had a transect layout. The number of chambers varied during the field campaign, but for the most part it consisted of seven flux chambers, with CO2-sensors in the

first and last chamber. The drifting chambers were attached behind the boat, since the boat drifted faster with the wind than the chambers, having roughly 9 m between the boat and a float to avoid chamber interference from the boat. From the float the distances between float and chambers as well as between chambers were the same as in the transect.

3.1.3 Sampling method and vial transfer

Open-air sample

Three full 60 ml syringes of sample were collected approximately 15 cm above the water surface. To transfer the sample to a vial, two 10 cm long Ø 5 mm (outer diameter) polyurethane extension tubes were used. These had a 3-way valve in one end, and a standard bore needle fitting in the other. Both extension tubes were fitted with needles. Both injection needles were inserted in a vial, with the 3-way valves on the extension tubes still closed. A syringe containing sample was fitted to one of the extension tubes, all 3-way valves opened, and the vial was flushed with the sample. The 3-way valves on the extension tubes were closed, and the syringe removed. The same procedure was done for the second and third syringe. When the third syringe had approximately (≥) 5 ml sample left, the 3-way valve on the outflow extension tube was closed, and the remaining sample forced into the vial, creating approximately 5 ml overpressure inside the vial. The overpressure was created to ensure sample integrity. In case of leakage, sample leaks out instead of air leaking in contaminating the sample. The 3-way valve on the other extension tube was closed, and the extension tubes removed from the vial.

Headspace sample

The headspace sample was treated in the same manner as the open-air sample with one exception, it was taken from the headspace of the flux chambers. This was enabled by the extension tube permanently connected to the flux chamber, fitted with a 3-way valve on the other end. The valve on the 60 ml syringe was connected to the 3-way valve on the extension tube from the chamber, the syringe was filled and emptied with headspace gas phase three times to create turbulence inside the chamber with the goal to sample a homogenous gas mixture from the chamber. The syringe was then filled a fourth time to collect the actual gas sample, both 3-way valves were closed to prevent contamination from the surrounding air, the syringe was removed, and the two remaining syringes were filled. Vial transfer was performed the same way as for the open-air sample.

Vial transfer

All samples were transferred to 20 ml glass vials, capped with a natural rubber cap and sealed airtight with an aluminum seal. The vials were sealed in the lab prior to sampling with lab air inside. The needles used were BD Microlance 0.4 x 13 mm injection needles. When inserting needles trough the rubber cap, the aim was always to hit the thicker part of the rubber cap, avoiding the thinner middle part. When removing syringes from the rubber caps, a pinch grip was held on the needle to prevent sample leakage by separation of syringe (or extension tube) from the needle. After removing needles from the rubber cap, a thumb grip was held firmly on the cap to give time for the rubber to seal the hole left from the removed needle. All syringes used were fitted with 3-way Luer-Lock valves to enable an airtight closing mechanism to the syringes.

Water temperature, coordinates, depth and air pressure

In addition to samples from the lake, background variables were collected. Some of these background variables were used in calculations (see Equation 4 and 6) to calculate the fluxes that are key to this study. To assist in collecting the background variables, the field boat was equipped with a Lowrance (LMS-520) sonar, which continuously logged water temperature approximately 30 cm below surface, GPS coordinates and water depth. Water temperature was recorded at the start and at the end of chamber deployment and was verified with a waterproof cooking thermometer. GPS coordinates were collected from Apple’s “Maps”-app on an iPhone. Depth was noted at the start and end of deployment, to avoid discrepancies within drift and transect chambers. Furthermore, air pressure needed for Equation 6 was collected from SMHI’s database from weather stations as close as possible to each sampling location.

3.1.4 CO

Sensor method

Chamber headspace CO2 concentration was measured with mini loggers according to Bastviken et al.

(2015). The sensors used were model K33 ELG from Senseair. It measures CO2 concentration in ppm,

temperature in °C and relative humidity in percent as well as logging date and time (Senseair, 2018). The sensors were modified according to Bastviken et al. (2015), waterproofing the otherwise exposed electrical board and pins and attaching cables for power supply and data access. For this study, the sensors were placed inside the chambers and set to measure CO2 concentration every sixth minute when

activated. Before the first measuring day, the sensors were calibrated outdoors. Data logs were retrieved from the sensors after each day to avoid overwriting of data due to a full memory.

3.2 Study sites

The lakes studied for this thesis were all what can be defined as large lakes (>10 km2) and are highlighted

in Figure 5. Since gas fluxes can be affected by several aspects, this section gives some information on how the lakes differ amongst each other. There were several reasons for these specific lakes being chosen for the study. The primary criterion was that they were large lakes, these being in proximity of Linköping University made it practically possible to visit the lakes and conduct sampling. Their different trophic states made the study more ample since studies have shown different fluxes for CH4 and CO2 depending

on the trophic levels in lakes (Beaulieu, DelSontro and Downing, 2019; Xiao et al., 2020). There are large differences in water depths in the three lakes, which according to Tranvik et al. (2009) further could impact results.

3.2.1 Glan

Glan is a natural 73 km2 large freshwater lake situated in Norrköping municipality. The lake is neither

shallow nor deep with its average depth of 9.9 m, and maximum depth of 22.8 m. The lake volume is 730 million m3 (SMHI, N.D.). Based on phosphorus concentrations from VISS (N.D.a) and trophic

classes in Carlson and Simpson (1996) it is a eutrophic lake.

Measurements in Glan were performed in the west to south-western parts of the lake (shown in Figure 6), close to the small community of Skärblacka. This location was chosen because it was close to the marina but not too close to be in proximity to boat traffic. The shores were dominated by forest with houses scattered throughout. Out of the four days on Glan, three days were either sunny or overcast and only one was rainy, the wind was generally calm. During the second period much algal bloom in the water was noticed in the lake, especially on August 6th.

Figure 6. Map showing the sampling locations for all sampling on Glan. The green and yellow dots represent

drifting chambers on deeper water. The lines show which way the drift was taking place each day. Blue and red dots represent transects (groups of five chambers) that are on shallow waters, close to shore.

3.2.2 Roxen

Roxen is a natural 95 km2 large freshwater lake in Linköping municipality. Despite its great area it is a

shallow lake with an average depth of 4.8 m, a maximum depth of 8 m and a lake volume of 458 million m3 (SMHI, N.D.). Data of phosphorus concentrations from VISS (N.D.b) in Roxen puts the lake as

mesotrophic in the classes from Carlson and Simpson (1996).

Measurements in Roxen were performed in the southeastern parts where, just like in Glan, the boat traffic was scarce and the site for the shallow transect was close to where the boat was launched (from the stream Stångån). The shallow transects were placed in a small bay where the water was shallow and especially during the second sampling period, there was a lot of plants and algae (see Figure 7).

Figure 7. Map showing the sampling locations for all sampling on Roxen. The green and yellow dots

represent drifting chambers on deeper water. The lines show which way the drift was taking place each day. Blue and red dots represent transects (groups of five chambers) that are on shallow waters, close to shore.

3.2.3 Vättern

Vättern is a natural, 1,886 km2 large freshwater lake that is part of several municipalities due to its sheer

size. For this study, the relevant part of Vättern is situated in Ödeshög municipality. Vättern is one of the deeper lakes in Sweden, with a maximum depth of 120 m, and an average depth of 40.8 m. Its volume amounts to 77,604 million m3 (SMHI, N.D.). Phosphorus concentration data from VISS (N.D.c) puts the

lake in the lower part of the oligotrophic class from Carlson and Simpson (1996), close to what some would define as hyper-oligotrophic.

The measurements in Vättern were made on the eastern side, in the middle part of the lake. The small populated area seen top right in Figure 8 is the port, Hästholmen. The weather was sunny and calm all days except the planned fourth day when waves and rough winds made sampling impossible. The lakes’ depth and the fact that Vättern is precipice made it difficult to pinpoint a shallow location where the transect would be less than 4 meters depth. Where the lakebed was visible, it consisted of large stone boulders and little sediment.

Figure 8. Map showing the sampling locations for all sampling on Vättern. The green and yellow dots

represent drifting chambers on deeper water. The lines show which way the drift was taking place each day. The second period, the starting location for the drift is missing since the GPS coordinates were incorrect. Blue and red dots represent transects (groups of five chambers) that are on shallow waters, close to shore.

3.3 Analysis method

Before analysis, all vials were tempered for at least one day in the same temperature as the analysis instrument and standards. Before sample analysis, vials were pressure equalized to remove the overpressure created in the field. This was done by injecting a needle with a water droplet in the needle hub to aurally indicate when equal pressure was achieved. All samples had an intact overpressure. CH4 concentration in gas samples was analyzed using a Gas Chromatograph (GC), of model 7890A with

a 1.8 m * 3.175 mm Porapak Q 80/100 Supelco column and a flame ionization detector, from Agilent Technologies, USA. The sample batches started and ended with a certified standard, and occasionally an open-air sample from just outside the lab (treated exactly as in field samples) was run to further verify that the instrument delivered results within reasonable intervals.Samples were injected with automatic injection using a 7697 headspace sampler attached to the 7890A. A single-point calibration was used with triple certified standards of (5000±100 ppm) as described by Natchimuthu et al. (2015).

3.4 Data treatment and flux calculations

3.4.1 Methane data

CH

data treatment

At first, the raw data of initial and final sample CH4 ppm levels were imported to an excel worksheet

together with the chamber area and volume data. Some unit conversions were made to fit the final equation for the flux calculation. Some of these were ppm to matm by multiplying current air pressure in atm with the conversion constant 9.869*10-9, which in turn was multiplied with the CH4 ppm value.

Current temperature measured in °C was converted to °K by adding 273.15 to the °C-value. Chamber volume measured in ml was converted to liters by dividing by 1000. Finally, the changes in CH4

concentration and time were calculated by subtraction of initial values from final values.

CH

Flux calculation

The CH4 flux was calculated in mmol m-2 d-1 as,

𝐹 =

A∗𝛥𝑡, (Eq 3)

where F is flux, A is chamber area in m2, Δt is the time change in days and Δn is the CH4 change in

mmol calculated as,

𝛥𝑛 =

𝑅∗𝑇 , (Eq 4)

where Δn is CH4 change in mmol, ΔP is the change in CH4 partial pressure in matm and V is flux chamber

volume in liters. Further, R is the gas constant (0.082056 L atm K-1mol-1) and T is initial water

3.4.2 Carbon dioxide data

CO

data treatment

The raw sensor data was imported to an excel worksheet using the free UIP5 software (www.senseair.com), where CO2, temperature and relative humidity were plotted against time in

separate series (see Figure 9). Next, all log entries outside the measurement period were removed to only include the correct data in the scatter plot. The function “SLOPE” was used to calculate the gradient coefficient in a regression line through four CO2 data-points. For the same four data points the function

“RSQ” was used to calculate the R2 value for said data points. This was done continuously for the whole

data series of each day.

To get a flux rate for extrapolation, a set of four data points was selected as the basis for calculation of slope, this slope value was then used as ΔS in Equation 6. To avoid subjective bias during this selection, a set framework was followed. The framework consisted of choosing the first set of data points— disregarding the first two measurements after deployment to avoid gas-stabilizing phase in the chamber—with an R2 value >0.9, that was also representative of the flux pattern for the following hour.

When an R2 value >0.9 was not achievable, a set of data points which’s slope the authors deemed

represented the pattern of the following hour of measurements was selected, while still striving for a high R2-value (more on this situation in the discussion). For drifting chambers, two slope values for each

sensor were chosen due to the large area covered by the chambers while for transect chambers only one slope value was selected as they were fixed in the same location.

Figure 9. Sensor data of CO2 concentration in one tenth ppm, chamber air temperature in °Celsius, and

chamber relative humidity in percent plotted against time for two days. The CO2 concentration is plotted

in 1/10 ppm to fit the same Y-axis as the temperature and relative humidity, this means that both dates the CO2 concentration starts at approximately 400 ppm. Example of chosen data points for calculating slope

value of CO2 shown highlighted in red. The chamber was deployed 11:00 and the first measurement point

CO

Flux calculation

The CO2 flux was calculated in mmol m-2 d-1 as,

𝐹 =

1000∗𝐴, (Eq 5)

where F is flux, A is chamber area in m2 and Δn is the CO2 change in mol d-1 calculated as,

𝛥𝑛 =

𝑃∗𝑅∗𝑇, (Eq 6)

where Δn is the CO2 change in mol d-1, ΔS is the slope rate in ppm d-1, V is the flux chamber volume in

liters. Further, P is the initial air pressure in atm, R is the gas constant (0.082056 L atm K-1mol-1) and T

is initial water temperature in °K.

3.4.3 Other data

Two other parameters were calculated to enable plotting results. Especially in the case of the transect chambers the time between these recordings of temperature could be several hours. To mediate changes and large differences in water temperature between the first and last temperature recording, a mean value between the two were used when plotting against water temperature.

Further the flux rate of CO2 was multiplied with 0.1 to scale down the values from 400-900 to 40-90.

This was done to allow plotting the CO2 levels on the same Y-axis as temperature in °C and relative

humidity in percent, which are both well below 400, and thus changes would not be visible in a graph with a Y-axis on a 400-900 scale.

3.5 Statistics

The statistical software IBM SPSS was used to analyze if there was a statistically significant difference in flux rate between and within the lakes. Since data was not normally distributed, non-parametric tests were used. Tests were run on each gas individually and a significance level of 0.05 was used.

Tests for statistical differences in flux rate were performed within each lake between the drifting chambers and the transect chambers. Tests were also performed within each lake between the two measuring periods. Since these were tests between two groups, the Mann-Whitney U-Test was used. This test assumes that all observations are independent of each other and of ordinal character. The null hypothesis is that the distributions of both groups is equal, and the alternate hypothesis is that the distributions are not equal.

Tests for difference in variance of flux rate were also performed within the drifting and transect chamber classes, but between the different lakes. Since it was a test between more than two groups, the Kruskal-Wallis test with post-hoc pairwise comparisons were used. Asymptotic significances for 2-sided tests were used, automatically adjusted in the software with the Bonferroni correction for multiple tests. This test assumes that measurements are collected at random, that the observations are independent and of ordinal character. The null hypothesis is that groups are from identical populations, the alternate hypothesis is that at least one group comes from another population than the others.

4. Results

4.1 Methane

4.1.1 Within lake variation

Glan

In Glan flux rates usually ranged between <0.10–0.90 mmol m-2 d-1. However, on 6th of August CH4

flux rates from drifting chambers were above any other recorded flux rates in the study, ranging 1.05– 1.78 mmol m-2 d-1 (highlighted in Figure 10). During this drifting measurement, water depth always

exceeded 7 m, and air pressure dropped below 1005 hPa for the first time in approximately 20 days. The total mean flux from Glan was 0.51 mmol m-2 d-1.

Temporally and spatially specific flux rates are compiled in Table 2. Figure 11 shows a pattern of spatial and temporal variability in Glan. Drifting chambers had higher flux rates than the fixed transect chambers, indicating spatial difference in Glan (Mann-Whitney U-Test; p=<0.001). The temporal variation is represented by the previously mentioned higher flux rates on August 6th. Excluding August

6th, there was no significant difference between the two periods (Mann-Whitney U-Test; p=0.015).

Roxen

In Roxen the flux rates ranged between <0.10–1.00 mmol m-2 d-1. The total mean flux from Roxen was

0.39 mmol m-2 d-1. Temporally and spatially specific flux rates are compiled in Table 2. Figure 11 shows

a pattern of spatial variation in Roxen; transect chambers consequently had higher flux rates than drifting chambers (Mann-Whitney U-Test; p=<0.001). In Roxen there was no significant temporal variation between the two periods (Mann-Whitney U-Test; p=0.921).

Vättern

Flux rates from Vättern were continuously close to zero throughout the whole study. Since the results from Vättern were close to zero, nothing could be said except that it always was <0.10 mmol m-2 d-1.

Figure 10. Scatter plot of CH4 flux rates (diffusion + ebullition) estimate against date. No difference

between drift and transect is made in this scatter plot. Highlighted in red are drifting measurements from August 6th, as they are higher than any other recorded flux rates.

Table 2. Table of estimated CH4 flux rate (diffusion + ebullition) for each lake, period (period details in

Table 1) and chamber types. “Missing” indicates chamber leakage.

Figure 11. Boxplot of estimated CH4 flux rate (diffusion + ebullition) within each lake between the

measuring periods, boxes clustered by drifting or transect chambers. See text above for details.

4.1.2 Between-lake variation

The CH4 flux results plotted for comparing the between-lake variation (Figure 12) indicated that fluxes

varied between lakes when compared within drift and transect groups. Test results (Kruskal-Wallis H-Tests; p=<0.001) showed that there was a significant difference between all three lakes within the two chamber groups (i.e., comparing drift with drift and transect with transect). Vättern showed close to no flux in both tests, unlike the other two lakes that had flux. Drift chamber flux rates were greater in Glan than in Roxen. Transect chamber flux rates were greater in Roxen than in Glan.

Figure 12. Boxplot of estimated CH4 flux rate (diffusion + ebullition) between lakes, boxes clustered by

drifting or transect chambers. See text above for details.

4.1.3 Flux correlation with environmental variables

Water temperature

When estimated CH4 flux rates were plotted against mean water temperature, a correlating pattern was

observed (see Figure 13). Flux rates from Vättern were, as mentioned above, close to zero at all times. However, the flux rate from Glan and Roxen seem to increase as the water temperature increases, as well as the distribution in flux rate between chambers.

Figure 13. Scatter plot of estimated CH4 flux rate (diffusion + ebullition) against mean water temperature.

4.2 Carbon dioxide

4.2.1 Within lake variations

Glan

Glan had the greatest flux rate variation of all lakes ranging from -144 to 200 mmol m-2 d-1 (this broad

variability was interestingly recorded the same day, 8th of August (Figure 14). For further information

about Glan, see Table 3). The drifting chambers and the transect chambers had no statistically significant difference between them (Mann-Whitney U-Test; p=0.462). When testing if there was a difference between period 1 and 2, no statistically significant difference could be established (Mann-Whitney U-Test; p=0.14).

Roxen

Roxen had lower flux rate variation than Glan with a minimum flux rate of -25.2 mmol m-2 d-1 and a

max flux rate of 25.2 mmol m-2 d-1 (see Table 3 and Figure 15). There was no statistically significant

difference between the drifting chambers and the transect chambers (Mann-Whitney U-Test; p=0.233). When testing if there was a difference in flux rates between Period 1 and 2, no statistically significant difference could be established (Mann-Whitney U-Test; p=0.588).

Vättern

Vättern had negative flux rates for both transect and drifting chambers as well as both periods (see Table 3 and Figure 15). Minimum and maximum flux rates ranged between 7.86 mmol m-2 d-1 and -10.8 mmol

m-2 d-1 respectively (note that the maximum flux rate was positive, but the overall mean was negative).

No statistically significant difference could be determined between the drifting and the transect chambers (Mann-Whitney U-Test; p=0.284). Between periods there was no statistically significant difference (Mann-Whitney U-Test; p= 0.934).

Table 3. Table of estimated CO2 flux rate for each period, lake and chamber types. “Missing” means that

flux rates were not possible to include in the study, either a sensor error occurred, or a chamber was leaking.

Figure 15. Boxplot of estimated CO2 flux within each lake between the measuring periods, boxes clustered

by drifting or transect chambers. On the x-axis the first and second measurement periods are shown. See text above for details.

4.2.2 Between-lake variation

CO2 flux rates were plotted to compare the between-lake variation (Figure 16). Looking at Figure 16,

flux rates and variability were greatest for Glan while Vättern had the lowest flux rates and variability. The Kruskal-Wallis H-Test was used to test differences between the lakes. The transect chambers and the drifting chambers were tested separately. The results for the tests only showed one statistically significant difference between the lakes: the drifting chambers between Roxen and Vättern (p=0.032). For the Kruskal-Wallis H-Test between Glan–Vättern and Glan–Roxen were thus not statistically significant (p=0.602 and p=0.427 respectively). That means that no statistically significant difference could be shown for transect chambers (the total grouped significance was p=0.474).

Figure 16. Boxplot of estimated CO2 flux between lakes in the same measuring period, boxes clustered by

drifting or transect chambers. See text above for details.

4.2.3 Flux correlation with environmental variables

Water temperature

Estimated CO2 flux was plotted with mean temperature for all lakes (Figure 17). The drift and transect

were split to investigate potential differences between these. Drifting chambers on Glan have a greater variability than on Roxen and Vättern. The transect chambers generally have a lower variability than the drifting chambers on all lakes but neither seem to correlate with temperature. This is not true for Vättern, which has close to zero variability and flux rate. For drift chambers there seems to be a small increase in variability as temperature rises.

Unexpected small-scale CO

flux variability

Upon studying the plotted raw data from sensors for picking which four values to be the basis of the slope calculations, unexpected distinct patterns of CO2 were visible. In the right half of Figure 18 a

gradual increase of CO2 levels from approximately 400 ppm to 950 ppm is clear and represents an

expected trend. However, in the left example the pattern is noticeably different. There is a distinct increase from approximately 450 ppm to 900 ppm, followed by roughly 20 minutes of insignificant changes, only to rapidly decrease again to approximately 500 ppm. Please note that the CO2

concentration results in ppm were multiplied with 0.10 to fit a shared Y-axis of 0-100.

Figure 18. Sensor data from a drifting flux chamber called “Sod 5” of CO2 concentration in one tenth ppm,

chamber air temperature in °Celsius, and chamber relative humidity in percent plotted against time for two days. The CO2 concentration is plotted in 1/10 ppm to fit the same Y-axis as the temperature and relative

humidity, this means that both dates the CO2 concentration starts at approximately 400 ppm. The data is

from the same sensor, fitted in the same chamber, in the same area on the same lake (Glan), at the same time during the day, in the same measurement period (Period 1), with just over a week between the two occasions.