2 methane emissions from a small wind shielded lake d...

Methane Emissions from a Small Wind Shielded Lake Determined by

Eddy Covariance, Flux Chambers, Anchored Funnels, and Boundary

Model Calculations: A Comparison

Carsten J. Schubert1,*, Torsten Diem1,b , Werner Eugster2

1 Swiss Federal Institute of Aquatic Science and Technology (Eawag), Dept. of Surface Waters - Research and Management, CH-6047 Kastanienbaum, Switzerland, Phone: 0041 58 765 2195, Fax: 0041 58 765 2168, email: [email protected]

2 ETH Zurich, Institute of Agricultural Sciences, CH-8092 Zurich, Switzerland

b Present address: University of St Andrews, School of Geography & Geosciences, St Andrews, KY16 9AL, Scotland, UK

of 31

ACS Paragon Plus Environment

Environmental Science & Technology

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Abstract

Lakes are large sources of methane, held to be responsible for 18% of the radiative

forcing, to the atmosphere. Periods of lake overturn (during fall/winter) are short and

therefore difficult to capture with field campaigns but potentially one of the most

important periods for methane emissions. We studied methane emissions using four

different methods, including eddy covariance measurements, floating chambers, anchored

funnels, and boundary model calculations. Whereas the first three methods agreed rather

well, boundary model estimates were 5-30 times lower leading to a strong

underestimation of methane fluxes from aquatic systems. These results show the

importance of ebullition as the most important flux pathway and the need for continuous

measurements with a large footprint covering also shallow parts of lakes. Although fluxes

were high, on average 4 mmol m-2 d-1 during the overturn period, water column microbial

methane oxidation removed 75% of the methane and only 25% of potential emissions

were released to the atmosphere. Hence, this study illustrates secondly the importance of

considering methane oxidation when estimating the flux of methane from lakes during

overturn periods.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Introduction

Atmospheric methane concentrations have increased from 0.800 ppm before

industrialization to a level of around 1.875 ppm today (1) and held responsible for 18% of

the radiative forcing (2). Whereas the biggest natural methane emissions stem from

wetlands, it was only recently suggested that a substantial part– between 6 and 16% of

natural methane emissions – might originate from lakes and other freshwater systems (3).

Conservative estimates assume that small lakes with surface areas smaller than 1 km2

account for only one third of the total emissions from lakes and freshwater systems. This

has now been called into question by more recent estimates of the global number of lakes

(4), which substantially exceeds previous estimates (5), and the findings that small lakes

have higher methane fluxes per unit area than larger lakes (6). Consequently, the amount

of methane emitted from small lakes (< 1 km2) might be much higher than previously

suggested.

We used 4 different methods including continuous eddy-covariance flux measurements

in combination with bi-weekly lake water profile measurements, floating chambers, and

anchored funnels to determine methane emissions during late season overturn. Eddy-

covariance flux measurements (7, 8) have so far not been used in natural aquatic systems

to estimate methane emissions. They are advantageous since they capture both diffusive

flux (normally estimated from surface-water methane concentrations) and ebullition

(normally determined using bubble capturing funnels). It is furthermore the only method

that allows continuous measurements over longer time periods (here 2.5 months).

In small lakes, methane stored in the hypolimnion, which can be emitted to the

atmosphere during lake turnover, may contribute up to 45% to the methane budget of

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

such lakes (9). This suggests a potential for large emissions during late season overturn.

Here we critically evaluate this methane emission potential with measurements from

Rotsee, a small holomictic lake (lakes which mix at least once a year (10)) in Switzerland

that tends to release its immense methane content, accumulated during the biologically

productive season, during mixing (11).

In lakes, the degradation of organic material (OM) is accomplished to a large extent via

methanogenesis (12), a process in which methane is produced by the fermentation of OM

or by carbon dioxide reduction (13). This methane is released from the sediment into the

water column, where it either diffuses or is transported via ebullition towards the water

surface, and is finally emitted to the atmosphere. During its passage through the water

column, methane concentrations are reduced by either aerobic or anaerobic oxidation

(11). However, if a chemocline (or any other physical barrier like a thermocline or

halocline) exists, the methane diffuses only very slowly through this barrier and can

accumulate in the hypolimnion during the productive season. Holomictic lakes such as

Rotsee begin to mix in fall or winter, when decreasing incident radiation and lower air

temperatures lead to a decrease in surface water temperatures and eventually to complete

mixing of the water column due to a density inversion (14).

Late season turnover is an important period for methane and carbon dioxide emissions

from lakes, since methane evasion rates by diffusion may be insignificant throughout

most of the year but very high during turnover periods (15-17).

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Materials and methods

Location

Lake Rotsee is a holomictic lake with a surface area of 0.5 km2 (2.5 km long and 200 m

wide). It is relatively wind shielded which led to wind speeds between 0.04 and 6.9 m s-1

in 1.25 m above the lake surface. During the measurement period the wind speed at 10 m

height was lower than 3.7 m s-1 during 1926 hours (1080 hours below 1 m s-1) and higher

than 3.7 m s-1 only during 69 hours.

Water temperature was between 14.7 and 7.3°C from 20 October to 1 December (see

Figure S1). The footprint area of the eddy covariance method is shown in Figure S2.

Eddy covariance flux measurements

Eddy-covariance flux measurements were conducted using a Gill/Solent 3-d ultrasonic

anemometer-thermometer (model R2A, Lymington, UK) and a Los Gatos Research

FMA-100 fast methane analyser (Mountain View, CA, USA) as described by Eugster &

Plüss (2009) (7). The sonic anemometer was mounted on a moored buoy 70 m from the

lake shore, and air (sucked from 1.25 m above lake surface) was guided to the methane

analyser using a Synflex 1300 tube (75 m) with an outer diameter of 10 mm. Turbulent

flow was maintained by a tri-scroll vacuum pump (BOC Edwards XDS-35i) with a flow

rate of 26.3 l min–1 (7). Data were recorded at 20.8 Hz resolution. Fluxes were computed

as covariances between vertical wind speed and concentration fluctuations after shifting

the methane time-series relative to the wind speed time-series to account for delays

caused by the tube and the methane analyser. The optimum time lag was determined with

a cross-correlation procedure for each 30-minute averaging interval with a search

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

window around the expected time delay (3.5–6.0 s). Fluxes were corrected for high-

frequency damping losses (18) to compensate for imperfect turbulent flow in the inlet and

in the vacuum cell of the methane analyser. A damping constant of 0.54 s–1 was used

based on least-squares fitting of a theoretical cospectrum to the measured cospectra. On

average, this correction increased measured fluxes by 35% (median 28%). Data quality

screening was done in the same way as done by Eugster et al. (2011) (8) for the data from

the Wohlensee run-of-river reservoir: 30-minute averages during which no clear cross-

correlation between vertical wind speed and CH4 concentration fluctuation was found

were removed and considered a flux that is not significantly different from zero. In

addition to the procedure discussed in detail by Eugster et al. (2011) (8) we used two

additional screening criteria which were specific for the Rotsee site: (a) 30-minute

averages during which the variance of CH4 concentration exceeded 0.4 ppm2, or (b)

during which the inclination angle of the streamlines of the wind field deviated more than

20° from the horizontal direction were removed. Since the sensors were mounted on a

floating buoy the second criterion removes both artefacts from strongly divergent flows

over the lake and conditions where the buoy was pushed downwind by the very rare

stronger winds.

Diffusive flux and ebullition

To estimate diffusive and ebullitive fluxes from the lake, two floating chambers (buoyed

buckets with a volume of 22 liter and a collecting area of 0.086 m2) were used.

The chamber sides were immersed a few centimeters into the water and had an outlet

tube connected to a three-way stopcock, which allowed the sampling of air with glass

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

syringes. At the beginning of each measurement a gas sample of 40 ml was taken and

injected into a glass bottle (30 ml) filled with NaCl-saturated water. While pushing the

gas into the bottle, the salt water was allowed to flow out of the bottle via a second

needle. After an hour of free floating, another glass bottle was filled with a second gas

sample as described above. To make sure the gas inside the chamber was mixed we

pumped the volume of the 50ml syringe several time in and out of the chamber. On most

sampling days two chambers were deployed twice for a total of 4 flux measurements,

except on sampling dates after the lake was mixed, when the chambers were only

deployed once, resulting in two flux measurements. The chambers were allowed to drift

over the lake surface for 0.8–2 hours on 19 days, in total for 72 hours (which compares to

6 m2 h).

Reported values represent the mean from one sample campaign. To derive a continuous

curve these mean values were added over time until the next sample campaign when new

mean values were obtained.

Additionally, two custom-made gas traps (funnels) with a collecting area of 0.79 m2 were

used to capture ebullition (19) from 24 October until 15 December. An air-tight cylinder

of known volume with a septum-lined cap was screwed onto the top of the funnel for gas

collection and sampling. The traps were deployed in an area upwind of the sonic

anemometer relative to the average wind direction using a moored buoy system that

allowed the funnel to hang upright approximately ~1 m below the water surface. When

deploying the funnels, care was taken to avoid catching gas bubbles that might be

liberated from the sediments upon anchoring. In total the area and time of the lake

covered by the funnels was 948 m2 h.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Gas samples taken from the cylinder every 2-3 days were transferred to a N2-filled serum

bottle for methane measurements. The amount of gas collected was determined by

measuring the height of the gas column in the cylinder and calculating the volume. Gas

samples were taken back to the lab and measured for methane concentrations as

described below. Methane emissions were calculated as follows:

tA

)(tCH)(tCH=)F(CH 44

4 12 (1)

where F[mg m–2 d–1] is the flux of methane from the water surface, CH4(tx) is the amount

of methane in the chamber [mg] at the beginning (x=1) and at the end (x=2) of the

measurement, A is the area of the floating chamber or funnel [m2] and t = t2-t1 is the

time [d] elapsed between the measurements.

Methane fluxes were calculated using the boundary layer model of Liss and Slater (1974)

(20):

F = 240 k (Cw - Ceq) (2)

The model estimates the air-water flux F [mg CH4 m-2 d-1] using the water saturation

concentration Ceq [M] according to Wiesenburg and Guinasso (1979) (21), the measured

water concentration Cw [M] of CH4 at 0.2 m water depth, the transfer velocity k [cm h-1]

and 240 is the unit conversion factor for the given units. For the calculation of the

transfer velocity k600 we used the bi-linear relationship given by Crusius and Wanninkhof

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

(2003) (22):for U10 < 3.7 m s-1 k600= 0 . 72U 10 cm h! 1

(3)

for U10 3.7 m s-1 k600= 4 .33 U 10! 13 .3 cm h

! 1

(4)

the relationship given by Cole & Caraco (1998) (23)

k600= 2 . 07+ 0 . 215 U101. 7

(5)

and the relationship given by MacIntyre et al. (2010) (24)

for a cooling lake k600= 2. 04 U 10+ 2. 0

(6)

for a heating lake k600= 1. 74 U10! 0. 15

(7)

where the lake was assumed to be cooling when air temperature was lower than the

surface lake temperature and heating for all other cases. Wind speeds at 10m were

obtained from the ones measured at 1.25m above lake surface with the sonic anemometer

and the correction given in equation (1) in Crusius & Wanninkhof (2003) (22), under the

assumption of neutral, stable boundary layer. The transfer velocity k600 was then

transformed into the transfer velocity k for methane using:

k = k600(Sc/600)c (8)

where Sc is the Schmidt number of the greenhouse gas (CH4) at water surface

temperature and c is -2/3 for U10 < 3.7 m s-1 and -1/2 for higher wind speeds (25).

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Methane concentrations and isotopic composition

Water samples were taken at 0.2, 2, 4, 6, 8, 10, 11, 11.5, 12, 12.5, 13, 13.5, 14, 14.5,

15, 15.5 m depth and transferred into 120 ml septum bottles directly after the Niskin

bottles came on board. The bottles were filled inserting a flexible tubing down to the

bottom of the bottle to avoid methane to escape by turbulent mixing, i.e., in the manner

commonly employed for Winkler titrations. The samples were poisoned using mercury

chloride, closed with a butyl stopper and sealed with an aluminium crimp. In the

laboratory a 20% headspace (He) volume was introduced and samples were equilibrated

overnight. Subsequently they were measured at a constant temperature with a gas

chromatograph (Agilent) equipped with a Carboxen 1010 column (30 m, Supelco) and a

flame ionization detector (FID). The oven temperature was 40°C. Methane standards

were made by dilution of pure CH4 (99.9%) and calibrated against commercial ones of

15ppm, 100ppm, 1000ppm and 1% (Scott, Supelco). Water concentrations were

determined according to McAuliffe (1971)(26) using equilibrium solubilities given by

Wiesenburg & Guinasso (1979)(21). Methane concentration in replicate samples varied

by <2%. Subsequently the methane concentrations in 0.2 m depth were used in the

boundary layer model of Liss & Slater (1974)(20) to calculate methane emission rates as

described above.

Methane isotopic composition was determined with a trace gas analyzer connected to a

mass spectrometer (GV Instruments). Notations are in the notation; i.e.:

13C = [(13C/12C)sample/( 13C/12C)standard] -1

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

The precision of the method was ± 0.7 ‰. We used isotopic measurements since changes

in methane isotopic composition towards heavier 13C values together with decreasing

dissolved methane concentrations can clearly distinguish oxidation in the water columns

from mixing of different water bodies.

Methane oxidation

Methane oxidation was estimated similar to the method described by Utsumi et al. (1998)

(27). Oxygen concentrations in the water column were measured with a SBE 19 CTD

(conductivity, temperature, depth) probe (Sea Bird Electronics) equipped with an oxygen

sensor (detection limit 1mol). At several depths (in total 65 samples at 20 days, see

Table S1) four replicate 125 ml bottles were filled with water to determine methane

concentrations as described above. Samples were taken with a flexible tubing reaching

the bottom of the bottle to avoid air intrusion into the sample. The sample depths were

chosen depending on the state of the oxycline, i.e., 1 to 5 samples were taken depending

on how well the oxic and anoxic water column was separated. In one sample biogenic

activity was stopped immediately with Cu(I)Cl, while the others were incubated in the

dark either at 7°C or at 20°C (whichever was closer to the water temperature). Every day

or every second day (for samples from the anoxic zone) one sample was poisoned with

HgCl2. After methane concentrations were measured as described above, methane

oxidation was calculated by the decrease in methane concentration over time. Methane

oxidation rates were then integrated over the whole water column and summed up

assuming rates to be the same from one sampling day to the following sampling day.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Results and discussion

Four different methods were employed to determine the magnitude of methane

emissions from the surface waters of Rotsee to the atmosphere. The first of these was

based on eddy-covariance flux measurements (7);Eugster, 2011 #19308}. A study carried

out with the same instruments by Tuzson et al. (2010)(28) confirmed that the eddy

covariance method correctly integrates over a footprint area with heterogeneous CH4

sources spaced 3-5 m apart and hence should also be able to correctly quantify ebullition

within the footprint area of the measurements (Figure S2).Methane concentrations from

14 October 2008 to 6 January 2009 varied between minimum values of 1.88±0.02 ppm

(close to the mean global methane concentration in the atmosphere) and maximum values

of 3.26±0.30 ppm. Methane fluxes determined using the eddy-covariance method are

shown in Figure 1b, with positive values denoting methane flux into the atmosphere.

Short-term positive and negative spikes must be considered noise that indicate the current

limits of system performance for direct CH4 flux measurements. Hence, we filtered 30-

minute data with a 5-point running average low-pass filter (bold line in Figure 1b) to

emphasis the signal that can be seen in our time series. It can clearly be seen that two

main emission events occurred during the three-month period; i.e., between 27 and 31

October and between 20 and 24 November. During these mixing events, responsible for

the main increments in the cumulative flux shown in Figure 1a, deep water containing

high concentrations of dissolved methane was mixed upwards towards the surface layer.

In the second method employed, methane emissions were calculated from surface water

methane concentrations using the boundary layer model of Liss & Slater (1974)(20).

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

In the third method, methane that escaped from the lake was collected in floating

chambers. Since these devices drifted freely distances of 50 to 400 m over the water

surface, travelling on most days across the width of the lake upwind of the sonic

anemometer and including both shallow littoral and deeper areas, these values were

considered representative of average methane emission values for the part of the lake

covered by the eddy-covariance measurement. This interpretation includes the experience

that disturbances heavily affect floating chamber measurements and hence fixing

chambers at fixed locations would most likely have failed to provide realistic flux

measurements, although they clearly would allow for an improved spatial representation

at the expense of considerable additional systematic errors in the overall flux

measurements.

Additionally, methane emission by ebullition was measured by locating two anchored

funnels connected to gas tight cylinders (19) above various sites. Sites from the shore

(shallow) towards the centre of the lake (deep), overall between 5 and 13 m deep, lying in

the footprint area of the eddy covariance method were chosen and included bubbling and

non-bubbling areas.

In Figure 2 methane emissions revealed by the different methods are presented. As would

be anticipated based on the variety of fluxes measured (diffusion only, diffusion plus

ebullition, ebullition only), the different methods yielded different results.

Cumulative methane effluxes derived from the eddy-covariance measurements (method

1), the floating chambers (method 3), and the funnels (method 4) are comparable: 4692,

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

4130, and 4558 mg m-2 (92.0, 81.0, and 89.4 mg m–2 d–1), respectively, for the period 27

October to 16 December, during which measurements were made using all four methods.

Fluxes derived using the formula of MacIntyre et al. (2010)(24) (830 mg m-2) were nearly

two times higher than those obtained using the formula of Cole & Caraco (1998) (23)

(460 mg m-2), which in turn were almost three times higher than those derived using the

formula of Crusius & Wanninkhof (2003) (22) (160 mg m-2). As our floating chamber

measurements neglected the decrease of diffusive flux due to increasing methane partial

pressure in the headspace we tried to correct for that by calculating diffusive fluxes with

a non-linear flux calculation. For this we used a set of 8 measurements which seemed to

be unaffected by ebullition (we chose a methane increase of less than 10 ppm in the

headspace as criteria) and calculated fluxes using the R (R Core Development Team,

2011)(29) library HMR (Pedersen, 2011)(30), which is based on a model by Hutchinson

& Mosier (1981)(31). This resulted in an average increase of fluxes by a factor of 1.5.

The diffusive flux we calculated with the boundary layer model represent between 3-18%

of the flux measured by eddy covariance. Therefore, assuming we underestimated the

diffusive fluxes measured with the floating chambers by a factor of 2/3 the corrected

emission of the floating chambers amount to 4200 - 4500 mg m-2 d-1.

Heat fluxes were calculated from the temperature measurement in the upper water

column (Figure S1) to reveal which process (convective turbulence or wind induced

turbulence) was the major factor during the experiment. On average heat flux was in a

normal range for lake surface water with 89 W m-2. Resulting convective turbulence had

average values of 1.8*10-8 W kg-1 (or m2 s-3 to compare with (24); values in our study

were between 1.2 and 5.8*10-8 W kg-1 at the lower end). Convective plume transport was

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

therefore on average 0.006 m s-1, whereas wind driven transport was between 0.0006 m s-

1 for wind speeds of 0.5 m s-1 and 0.009 m s-1 for wind speeds of 7 m s-1. This clearly

shows that above 4.3 m s-1 wind speed the turbulence induced by wind was the dominant

factor, whereas below this value the convective turbulence was the relevant emission

factor. This additionally to the discussion below explains why the emission estimates

with the boundary model give relatively low values compared to the other methods, since

at the time of our experiment only during 31 hours wind speeds were higher than 4.3 m s-

1 compared to 1963 hours were convective turbulence prevailed.

Our results confirm that effluxes derived from the boundary layer model may strongly

underestimate methane fluxes (5-30 fold compared to methods 1, 3, and 4) when

ebullition is present and convective turbulence is dominant in the lake (19). Thus we can

conclude that the eddy-covariance method, floating chambers, and a combination of

ebullition measurements with additional diffusive flux calculation using a turbulent

boundary layer model (e.g. Crusius & Wanninkhof, 2003 or Cole & Caraco, 1998) (22,

23) or k600 values derived from eddy-covariance measurements (e.g. (24)) equally

represent lake emissions. It has to be noted that the uncertainties of the funnel and

floating chamber fluxes are most likely higher than the ones of the eddy covariance

measurements which is clearly the method of choice. This is mainly due to the fact that

eddy covariance would detect diffusive plus ebullition flux and that it could be set up to

measure continuously over a longer time span. Our results confirm that ebullition is the

most important emission pathway (3, 19). This is in agreement with a very recent report

in which it was noted that since ebullition is not generally included into aquatic emission

pathways, the emissions by inland waters are most likely underestimated (32). Hence, we

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

strongly recommend either eddy covariance or a combination of ebullition and diffusive

flux measurements for estimating methane effluxes from aquatic systems. Still, a word of

caution needs to address potential problems with spatial representation of a whole lake

with the methods used here: during manual sampling the visual detection of bubbles at

the lake surface helps to detect ebullition hot spots. However, ebullition hot spots that are

active during the absence of researchers on the lake (actually most of the time) may have

been missed by our approach and may need a correction in future. Follow-up studies,

both with respect to the experimental design and the procedure how to extrapolate

existing measurements to the entire lake surface are necessary.

Looking at the eddy-covariance flux measurements (Fig. 1) it is obvious that, despite

these large effluxes during the two main turnover events, the total amount of methane

released from Rotsee was only 5.4 g CH4 m-2 during winter overturn (84 days). On the

one hand this is high, i.e., on the order of 50% of daily emissions that have recently been

reported from tropical reservoirs (33). On the other hand, this is rather low compared to

the potential emissions of 21.3 g CH4 m-2 that could be postulated simply by taking the

total amount of methane dissolved in the hypolimnion into account (i.e., neglecting new

methane production).

What is happening in the water column during these events? Figure S1 (supplement)

shows the oxygen concentration profile on 20 October, with around 11 mg O2 l–1 in the

uppermost 9 m. The lack of oxygen below this depth indicates the existence of anoxic

conditions in the hypolimnion. The oxygen measurements from 20 October to 1

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

December clearly show the mixing of the water column with oxygen penetrating more

and more deeply into the hypolimnion. This is especially evident between 14 and 26

November, when we measured strong emissions with the eddy-covariance method, and

the oxycline moved from above 11 m to below 14 m depth. Figure 3 shows the methane

concentrations in the water column before and during this mixing event (20 and 24

November with concentrations in the bottom water of up to 16 mg l-1 (1 x 103 M).

Higher dissolved methane concentrations towards the sediment identify the sediments to

be the source of the methane. On 17 November, methane concentrations in the uppermost

10 m were in the 3.5 - 11.2 g l-1 (0.2 – 0.7 M) range. After a strong mixing event on 21

November, which moved the oxycline about 3 m downwards (Fig. S1 supplement),

methane concentrations in the upper 10 m were 15- to 20-fold higher (73-170 g l-1 or

4.6-10.6 M). After 3 days, however, these high concentrations in the surface waters had

already strongly decreased to 15-23 g l-1 (1-1.5 M), indicating strong oxidation of

methane in the oxic zone, which on 24 November already included the top 14 m of the

lake. Measured methane oxidation rates (Fig. S4) and isotope analysis (Fig. 3) give strong

support for methane oxidation to be largely responsible for this strong decrease and that

diffusion is of less importance (a diffusion-driven profile would show decreasing values

towards the surface). This is a clear indication that oxidation in the mixed layer of the

lake is the primary removal process for methane, and only a small fraction is emitted.

Isotope analysis yields further strong support for this interpretation (Fig. 3). The isotopic

composition of the carbon in the dissolved methane during the mixing event on 21

November (13C = -66 ‰ on average) differed drastically from that of the methane in the

oxic zone before the event ( 13C = -30 ‰ on average), indicating strongly that this

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

methane originates from the hypolimnion. However, by 24 November, 3 days after the

mixing event, the 13C of the methane had not just returned to the value it had had on 17

November, but had even become heavier by ~15 ‰. The drastic change from a 13C

value of -66‰ to -15 ‰ (fractionation factor =0.948) clearly indicates that methane was

oxidized by methanotrophs (microorganisms that consume methane), since

microbiologically mediated oxidation of methane is accompanied by a strong isotopic

fractionation, leaving heavier methane behind (11, 34). We measured average methane

oxidation rates of 26 ± 43 M CH4 l-1 d-1 in the anoxic zone/chemocline and 1.0 ± 2.3

M CH4 l-1 d-1 in the oxic zone (Fig. S3 supplement). A peak value of 12 M CH4 l

-1 d-1

was measured in the oxic zone during the second strong mixing event on 23 November.

This provides evidence for an existing methanotrophic community (mainly

Methylomonas and Methylobacter (11) in the oxic zone capable of oxidizing the methane

which occurs in very high concentrations in the epilimnion during lake mixing. That

methanotrophic communities react very quickly to increased methane concentration has

recently been shown in the Gulf of Mexico oil spill (35). The generally higher oxidation

rates with a peak value in the anoxic zone/chemocline of 225 M CH4 l-1 d-1 provide

clear evidence that anaerobic oxidation is important. However, the microorganisms that

mediate the anaerobic oxidation of methane in lakes are still unknown.

In Figure 4 cumulative values for both methane oxidation (7 g m-2 in the oxic water layer

and 33 g m-2 in the whole water column) and methane effluxes (5.4 g m-2) are shown. It

becomes obvious that rates of methane oxidation exceed rates of methane effluxes by a

factor of about 6 owing to the very effective microbiological barrier that hinders larger

methane effluxes to the atmosphere. Hence, true methane emissions are only 25% of

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

what could potentially be released if all stored methane in the hypolimnion was to be

emitted without microbial oxidation (i.e., 2.5 t of 10 t CH4 are emitted during late season

turnover). These values are in good agreement with other investigations; e.g. from a

Japanese lake, where 74% of the methane was oxidized (27), and from a Finnish lake,

where 83-88% of the methane was oxidized (36).

With 75% oxidation occurring in the water column and a greenhouse gas forcing factor of

25 CO2-equivalents per kg of CH4 (IPCC 2007) (2), this means that each mass unit of

methane produced in the hypolimnion represents only 6.25 CO2-equivalents at the lake

surface; i.e., only 25% of the greenhouse gas forcing that would be erroneously assumed

from the gas production in sediment cores if oxidation in the lake was neglected. Hence,

our findings have strong implications for the recently proposed role that lakes, and to a

smaller extent reservoirs, might play not only in the global methane budget (37) (3) but

also in the continental carbon sink estimations (32). Although all lakes (n=4) and

reservoirs (n=11) Diem et al. (2008) investigated in Switzerland are methane emitters

(38), they were able to show that the amount of methane released over the year by these

lakes, including that released during late season turnover, is much smaller than it would

be if all the methane stored in the hypolimnion were to be released to the atmosphere

without microbial interaction. This reduces the importance of storage in the emission

estimates (e.g. Michmerhuizen et al. (1996)(39) only assumed an oxidation of 1-7% of

the storage methane) and counteracts the possibly higher number of small lakes as

pointed out recently (4) . In return, it implies that the huge anthropogenic methane

emissions from rice fields, cattle, and energy production are even more important as

greenhouse gas sources, and are worth reducing in the future if we wish to limit the

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

increase in global surface air temperature due to greenhouse gas production to 2°C.

Finally, we showed that methane emission can yield similar results, when quantified by

different methods and over different time frames. Whether this conclusion is valid across

different lakes and different times of the year, or if different methods are more suitable

for certain conditions needs to be further investigated. Nevertheless, it will be interesting

to see in how much future emission estimates using more sophisticated measurements

including eddy covariance will increase existing greenhouse gas emission estimates from

continental aquatic systems.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Acknowledgements

We would like to thank Beat Müller, Michael Schurter, and Oliver Scheidegger for their

help during field work and BM for providing the bathymetry data. We also thank Johny

Wüest for his support in calculating heat fluxes as well as convective and wind induced

turbulence. This project was supported by an ETH Scientific Equipment grant, the

Competence Centre for Environment and Sustainability (CCES), and funds from the

Swiss Federal Institute of Aquatic Science and Technology (Eawag).

Supporting Information Available. This information is available free of charge via the

Internet at http://pubs.acs.org/.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://pubs.acs.org/

Figures

Figure 1

Cumulative methane flux (a) and methane flux (b) during lake turnover (14 October 2008

to 6 January 2009) estimated from eddy covariance measurements. The thick line in panel

(b) shows 30-minute flux averages, whereas the bold line shows the 2.5 hour low-pass

filtered values. Two major emission events can be noted between 27 October and 31

October and between 20 and 24 November. Cumulative methane fluxes amount to 5.4 g

m-2. DOY (day of the year).

Figure 2

Methane emissions measured during lake turnover (14 October 2008 to 6 January 2009)

by different methods including the eddy covariance method, the boundary layer model

(Liss & Slater 1974) (20) using transfer velocity k600 calculated with the bi-linear

relationship given by Crusius & Wanninkhof (2003)(22),the relationship given by Cole &

Caraco (1998) (23), the relationship by Mac Intyre et al. (2010) (24) chambers floating

freely over the lake surface, and funnels collecting methane bubbles. Several chamber

and funnel measurements over one day were averaged to give a daily mean. Those mean

values were used to interpolate CH4 flux values until the next sampling date to obtain a

continuous curve. Orange and violet squares indicate when floating chamber and funnel

samples were taken, respectively. Whereas methane emissions estimated with the eddy

covariance, the funnels, and floating chambers agree very well,values calculated with the

boundary layer model are much lower.

Figure 3

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Methane concentrations (a) and methane carbon isotopes (b) measured on three different

days during a major emission event between 17 November (DOY 322) and 24 November

(DOY 329). On 17 November methane concentrations are low in the epilimnion, an

increase to high values on a mixing event on 21 November (DOY 326) can be noted,

however, three days later methane concentrations were already much lower again. The

carbon isotope values indicate strongly oxidized methane on 17 and 24 November and

original source methane (not oxidized) from the hypolimnion on 21 November during the

mixing event.

Figure 4 Cumulative methane efflux to the atmosphere and oxidation in the oxic and anoxic water

column of Rotsee. Methane oxidation rates were assumed to be the same from one

sampling date to the other (altogether 20 days, see Table S1). It can be noted that the

combined aerobic and anaerobic oxidation of methane in the water column clearly

exceeds methane emission to the atmosphere by a factor of 7 to 8.

(1) Dlugokencky, E. J.; Bruhwiler, L.; White, J. W. C.; Emmons, L. K.; Novelli, P. C.; Montzka, S. A.; Masarie, K. A.; Lang, P. M.; Crotwell, A. M.; Miller, J. B.; Gatti, L. V., Observational constraints on recent increases in the atmospheric CH4 burden. Geophysical Research Letters 2009, 36 (18), L18803. (2) Forster, P.; Ramaswamy, V.; Artaxo, P.; Berntsen, T.; Betts, R.; Fahey, D. W.; Haywood, J.; Lean, J.; Lowe, D. C.; Myhre, G.; Nganga, J.; Prinn, R.; Raga, G.; Schulz, M.; Dorland, R. V., Changes in Atmospheric Constituents and in Radiative Forcing. In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Solomon, S.; Qin, D.; Manning, M.; Chen, Z.; Marquis, M.; Averyt, K. B.; Tignor, M.; Miller, H. L., Eds. Cambridge University Press:Cambridge, UK, and New York, NY, 2007.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

(3) Bastviken, D.; Cole, J.; Pace, M.; Tranvik, L., Methane emissions from lakes: Dependence of lake characteristics, two regional assessments, and a global estimate. Global Biogeochemical Cycles 2004, 18, (4), 1-12. (4) Downing, J. A.; Prairie, Y. T.; Cole, J. J.; Duarte, C. M.; Tranvik, L. J.; Striegl, R. G.; McDowell, W. H.; Kortelainen, P.; Caraco, N. F.; Melack, J. M.; Middelburg, J. J., The global abundance and size distribution of lakes, ponds, and impoundments. Limnology and Oceanography 2006, 51, (5), 2388-2397. (5) Kalff, J., Limnology. Prentice Hall: New York, 2002; p 608. (6) Juutinen, S.; Rantakari, M.; Kortelainen, P.; Huttunen, J. T.; Larmola, T.; Alm, J.; Silvola, J.; Martikainen, P. J., Methane dynamics in different boreal lake types. Biogeosciences 2009, 6, (2), 209-223. (7) Eugster, W.; Pluess, P., A fault-tolerant eddy covariance system for measuring CH(4) fluxes. Agricultural and Forest Meteorology 2010, 150, (6), 841-851. (8) Eugster, W.; DelSontro, T.; Sobek, S., Eddy covariance flux measurements confirm extreme CH(4) emissions from a Swiss hydropower reservoir and resolve their short-term variability. Biogeosciences 2011, 8, (9), 2815-2831. (9) Bastviken, D.; Cole, J. J.; Pace, M. L.; Van de Bogert, M. C., Fates of methane from different lake habitats: Connecting whole-lake budgets and CH4 emissions. Journal of Geophysical Research-Biogeosciences 2008, 113, (G2). (10) Wetzel, R. G., Limnology. Saunders College Publishing: Philadelphia, 1983; p 767. (11) Schubert, C.; Lucas, F.; Durisch-Kaiser, E.; Stierli, R.; Diem, T.; Scheidegger, O.; Vazquez, F.; Müller, B., Oxidation and emission of methane in a monomictic lake (Rotsee, Switzerland). Aquatic Sciences - Research Across Boundaries 2010, 1-12. (12) Conrad, R., Control of methane production in terrestrial ecosystems. In Exchange of Trace Gases between Terrestrial Ecosystems and the Atmosphere, Andreae, M. O.; Schimel, D. S., Eds. John Wiley & Sons Ltd: Chichester, 1989; Vol. 47, pp 39-58. (13) Whiticar, M. J.; Faber, E.; Schoell, M., Biogenic methane formation in marine and freshwater environments; CO2 reduction vs. acetate fermentation; isotope evidence. Geochimica et Cosmochimica Acta. 1986, 50, (5), 693-709. (14) Lerman, A.; Imboden, D.; Gat, J., Physics and Chemistry of Lakes. Springer: Berlin, 1995. (15) Riera, J. L.; Schindler, J. E.; Kratz, T. K., Seasonal dynamics of carbon dioxide and methane in two clear-water lakes and two bog lakes in northern Wisconsin, USA. Canadian Journal of Fisheries and Aquatic Sciences 1999, 56, (2), 265-274. (16) Kankaala, P.; Huotari, J.; Peltomaa, E.; Saloranta, T.; Ojala, A., Methanotrophic activity in relation to methane efflux and total heterotrophic bacterial production in a stratified, humic, boreal lake. Limnology and Oceanography 2006, 51, (2), 1195-1204. (17) Kankaala, P.; Taipale, S.; Nykanen, H.; Jones, R. I., Oxidation, efflux, and isotopic fractionation of methane during autumnal turnover in a polyhumic, boreal lake. Journal of Geophysical Research-Biogeosciences 2007, 112, (G2). (18) Eugster, W.; Senn, W., A cospectral correction model for measurement of turbulent NO2 flux. Boundary-Layer Meteorology 1995, 74, (4), 321-340. (19) Delsontro, T.; McGinnis, D. F.; Sobek, S.; Ostrovsky, I.; Wehrli, B., Extreme methane emissions from a Swiss hydropower reservoir: contribution from bubbling sediments. Environmental Science & Technology 2010, 44, (7), 2419-2425.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

(20) Liss, P.; Slater, P., Flux of gases across the air-sea interface. Nature 1974, 247, 181-184. (21) Wiesenburg, D.; Guinasso, N., Equilibrium solubilities of methane, carbon-monoxide, and hydrogen in water and sea water. Journal of Chemical and Engineering Data 1979, 24, 356-360. (22) Crusius, J.; Wanninkhof, R., Gas transfer velocities measured at low wind speed over a lake. Limnology and Oceanography 2003, 48, (3), 1010-1017. (23) Cole, J. J.; Caraco, N. F., Atmospheric exchange of carbon dioxide in a low-wind oligotrophic lake measured by the addition of SF6. Limnology and Oceanography 1998, 43, (4), 647-656. (24) MacIntyre, S.; Jonsson, A.; Jansson, M.; Aberg, J.; Turney, D. E.; Miller, S. D., Buoyancy flux, turbulence, and the gas transfer coefficient in a stratified lake. Geophysical Research Letters 2010, 37, (24), L24604. (25) Liss, P.; Merlivat, L., Air-sea gas exchange rates: Introduction and synthesis In Dordrecht, Holland, 1986; p 113-127. (26) McAuliffe, C., GC determination of solutes by multiple phase equilibration. Chemical Technology 1971, 46-51 (27) Utsumi, M.; Nojiri, Y.; Nakamura, T.; Nozawa, T.; Otsuki, A.; Seki, H., Oxidation of dissolved methane in a eutrophic, shallow lake: Lake Kasumigaura, Japan. Limnology and Oceanography 1998, 43, (3), 471-480. (28) Tuzson, B.; Hiller, R. V.; Zeyer, K.; Eugster, W.; Neftel, A.; Ammann, C.; Emmenegger, L., Field intercomparison of two optical analyzers for CH4 eddy covariance flux measurements. Atmospheric Measurement Technics 2010, 3, (6), 1519-1531. (29) R_Development_Core_Team, R: A language and environment for statistical computing. Vienna, Austria, 2011. (30) Pedersen, A. R., HMR: Flux estimation with static chamber data. R package version 0.3.1. http://CRAN.R-project.org/package=HMR 2011. (31) Hutchinson, G. L.; Mosier, A. R., Improved soil cover method for field measurement of nitrous oxide fluxes. Soil Sci. Soc. Am. J 1981, 45, 311-316. (32) Bastviken, D.; Tranvik, L. J.; Downing, J. A.; Crill, P. M.; Enrich-Prast, A., Freshwater methane emissions offset the continental carbon sink. Science 2011, 331, (6013), 50-50. (33) Bastviken, D.; Santoro, A. L.; Marotta, H.; Pinho, L. Q.; Calheiros, D. F.; Crill, P.; Enrich-Prast, A., Methane emissions from Pantanal, South America, during the low water season: toward more comprehensive sampling. Environmental Science & Technology 2010, 44, (14), 5450-5455. (34) Barker, J. F.; Fritz, P., Carbon isotope fractionation during microbial methane oxidation. Nature 1981, 293, (5830), 289-291. (35) Kessler, J. D.; Valentine, D. L.; Redmond, M. C.; Du, M.; Chan, E. W.; Mendes, S. D.; Quiroz, E. W.; Villanueva, C. J.; Shusta, S. S.; Werra, L. M.; Yvon-Lewis, S. A.; Weber, T. C., A persistent oxygen anomaly reveals the fate of spilled methane in the deep Gulf of Mexico. Science 2011, 331, (6015), 312-315.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://cran.r-project.org/package=HMR

(36) Kankaala, P.; Eller, G.; Jones, R. I., Could bacterivorous zooplankton affect lake pelagic methanotrophic activity? Fundamental and Applied Limnology 2007, 169, (3), 203-209. (37) St Louis, V. L.; Kelly, C. A.; Duchemin, E.; Rudd, J. W. M.; Rosenberg, D. M., Reservoir surfaces as sources of greenhouse gases to the atmosphere: A global estimate. Bioscience 2000, 50, (9), 766-775. (38) Diem, T.; Koch, S.; Schwarzenbach, S.; Wehrli, B.; Schubert, C. J., Greenhouse gas emissions (CO2, CH4 and N2O) from perialpine and alpine hydropower reservoirs. Biogeosciences Discussions 2008, 5, 3699-3736. (39) Michmerhuizen, C. M.; Striegl, R. G.; McDonald, M. E., Potential methane emission from north-temperate lakes following ice melt. Limnology and Oceanography 1996, 41, (5), 985-991.

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

0

1

2

3

4

5

6

Cum

ul. C

H4 fl

ux (g

m-2)

6thOct 26thOct 15thNov 5thDec 25thDec 14thJan

280 300 320 340 360 380-10

-5

0

5

10

15

CH

4 flux

(µg

m-2s-1

)

DOY

Fig. 1 of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

280 300 320 340 360 380

0

2000

4000

6000 C&C C&WMacIntyre Floating ch. Funnels Eddy Cov.

C

H4

efflu

x (m

g m

-2)

DOY

Fig. 2 of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

-80 -70 -60 -50 -40 -30 -20 -10 016

14

12

10

8

6

4

2

0

!

Dep

th (m

)

δ"#$%&'()*&! +,! -./! 01234

5

100 1000 10000 100000 100000016

14

12

10

8

6

4

2

0

!

! 67-&85&9! ":'(! 67-&85&9! ;"'(! 67-&85&9! ;<'(

Dep

th (m

)

%&'()*&! *87=! =>"

)

?@A/! # of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

280 300 320 340 360 3800

10000

20000

30000

40000

Oxidation (oxic) Emission (EC) Oxidation (anoxic)

Sum

of m

etha

ne e

mis

sion

/oxi

datio

n (m

g m

-2)

DOY

Fig. 4

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Methane in aquatic systems

Met Station

Eddy Correlation (EC) sensor

Los Gatos CH4 sensor

CH4 sensor pump

MAIOLICA (CCES project)

Werner Eugster ETH Institute of Grassland Science

CH4 sensor box

High-resolution direct- absorption spectroscopy

280 300 320 340 360 380

0

2000

4000

6000 C&C C&WMacIntyre Floating ch. Funnels Eddy Cov.

CH4 e

fflux

(mg

m-2

)

DOY

Fig. 2

of 31



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

2 methane emissions from a small wind shielded lake d...

Documents