Temporal Variation of Ecosystem Scale Methane Emission
From a Boreal Fen in Relation to Temperature, Water
Table Position, and Carbon Dioxide Fluxes
Janne Rinne1 , Eeva-Stiina Tuittila2 , Olli Peltola3 , Xuefei Li3, Maarit Raivonen3,
Pavel Alekseychik3, Sami Haapanala3 , Mari Pihlatie3,4 , Mika Aurela5, Ivan Mammarella3 ,
and Timo Vesala3,6
1Department of Physical Geography and Ecosystem Science, Lund University, Lund, Sweden, 2School of Forest Sciences,
University of Eastern Finland, Joensuu, Finland, 3Institute for Atmospheric and Earth System Research/Physics, Faculty of
Science, University of Helsinki, Helsinki, Finland, 4Environmental Soil Science, Department of Agricultural Sciences, Faculty
of Agriculture and Forestry, University of Helsinki, Helsinki, Finland, 5Finnish Meteorological Institute, Helsinki, Finland,
6Institute for Atmospheric and Earth System Research/Forest Sciences, Faculty of Agriculture and Forestry, University of
Helsinki, Helsinki, Finland
Abstract We have analyzed decade-long methane flux data set from a boreal fen, Siikaneva, together
with data on environmental parameters and carbon dioxide exchange. The methane flux showed
seasonal cycle but no systematic diel cycle. The highest fluxes were observed in July–August with average
value of 73 nmol m
2 s
1. Wintertime fluxes were small but positive, with January–March average of
6.7 nmol m
2 s
1. Daily average methane emission correlated best with peat temperatures at 20–35 cm
depths. The second highest correlation was with gross primary production (GPP). The best correspondence
between emission algorithm and measured fluxes was found for a variable-slope generalized linear model
(r2 = 0.89) with peat temperature at 35 cm depth and GPP as explanatory variables, slopes varying between
years. The homogeneity of slope approach indicated that seasonal variation explained 79% of the sum of
squares variation of daily average methane emission, the interannual variation in explanatory factors 7.0%,
functional change 5.3%, and random variation 9.1%. Significant correlation between interannual variability of
growing season methane emission and that of GPP indicates that on interannual time scales GPP controls
methane emission variability, crucially for development of process-based methane emission models.
Annual methane emission ranged from 6.0 to 14 gC m
2 and was 2.7 ± 0.4% of annual GPP. Over 10-year
period methane emission was 18% of net ecosystem exchange as carbon. The weak relation of methane
emission to water table position indicates that space-to-time analogy, used to extrapolate spatial chamber
data in time, may not be applicable in seasonal time scales.
Plain Language Summary Methane emission from a boreal wetland was measured over one
decade. Methane emission shows strong seasonal cycle, with highest emission in late summer and lowest
emission during winter. No diel cycle was observed. The methane emission is an important part of the
carbon balance of the wetland as 18% of carbon taken up as carbon dioxide was emitted back into
atmosphere as methane. The seasonal cycle of the emission was controlled first by peat temperature and
second by ecosystem photosynthesis. The interannual variability of methane emission was more related to
photosynthesis. A large part of the interannual variability remained unexplained by the measured
environmental parameters.
1. Introduction
Methane is the second most important greenhouse gas in contributing to anthropogenic climate change
(IPCC, 2013). Its existence and part of origins have been known since observations and experiments by
Volta (1777) and its strong infrared absorption properties since experiments by Tyndall (1862). There exist
several biogenic and anthropogenic sources of methane into the atmosphere, the most important biogenic
one being wetlands (Ciais et al., 2013). In wetlands, the high water-table maintains anaerobic conditions
allowing only anaerobic organisms. In northern peat accumulating wetlands, that is, peatlands, organic soils
and input of fresh hydrocarbons as root exudates from vegetation provide ideal conditions for methane-
producing archaea. The net methane emission from wetland ecosystems is a balance between methane
RINNE ET AL. 1087
Global Biogeochemical Cycles
RESEARCH ARTICLE
10.1029/2017GB005747
Key Points:
• Methane emission was insensitive to
water table variations observed
during decade-long period
• Soil temperature was the dominant
driver of methane emission during
shorter periods of time (<1 year),
while interannual variation was more
related to variation in GPP
• Seventy-nine percent of the
variability of daily average methane
emission is due to seasonal change of
drivers, 7.0% is due to interannual
variation of drivers, and 5.3% is due
to functional change
Supporting Information:
• Supporting Information S1
Correspondence to:
J. Rinne,
janne.rinne@nateko.lu.se
Citation:
Rinne, J., Tuittila, E.-S., Peltola, O., Li, X.,
Raivonen, M., Alekseychik, P., et al.
(2018). Temporal variation of ecosystem
scale methane emission from a boreal
fen in relation to temperature, water
table position, and carbon dioxide
fluxes. Global Biogeochemical Cycles, 32,
1087–1106. https://doi.org/10.1029/
2017GB005747
Received 20 JUN 2017
Accepted 16 JUN 2018
Accepted article online 27 JUN 2018
Published online 20 JUL 2018
©2018. The Authors.
This is an open access article under the
terms of the Creative Commons
Attribution-NonCommercial-NoDerivs
License, which permits use and distri-
bution in any medium, provided the
original work is properly cited, the use is
non-commercial and no modifications
or adaptations are made.
produced in anaerobic conditions by archaea and methane consumed by methane oxidizing microorgan-
isms in the aerobic peat layers.
Fens are estimated to make up nearly half of the total global peatland area (Aselmann & Crutzen, 1989;
Lamers et al., 2015) and their contribution to methane emission from peatlands to be over 80% (Aselmann
& Crutzen, 1989). Especially in the northern peatlands, higher methane emissions have been measured
from fens than from ombrotrophic bogs (Turetsky et al., 2014). Hence, understanding methane emission
dynamics from fen ecosystems is crucial for understanding the methane emission from northern peatlands.
Furthermore, as current bogs have transitioned from fens to bogs at some point of their history, understand-
ing the Holocene greenhouse gas balance and climatic effect of peatland requires understanding of the fen
emission dynamics (Mathijssen et al., 2016).
The methane emissions from wetlands have been intensively studied by chamber techniques (e.g., Bubier
et al., 2005; Riutta et al., 2007; Ström et al., 2015; Whiting et al., 1991; Whiting & Chanton, 1992), and more
recently also by eddy covariance method, measuring the ecosystem scale gas exchange (e.g., Brown et al.,
2014; Fortuniak et al., 2017; Hargreaves et al., 2001; Hommeltenberg et al., 2014; Jackowicz-Korczyński
et al., 2010; Kim, Verma, & Billesbach, 1998; Kim, Verma, Billesbach, & Clement, 1998; Kowalska et al., 2013;
Mikhaylov et al., 2015; Nadeau et al., 2013; Rinne et al., 2007). While chamber measurements provide insight
to many processes and to microtopography-scale spatial variability, they usually lack the temporal resolution
of eddy covariance data. Also, the deployment of chamber systems to obtain long-term measurements with-
out disturbance to studied ecosystem, such as changes in snow accumulation inside collar, warming effect of
many automated chamber systems, is challenging. The eddy covariance method (e.g., Aubinet et al., 2012),
which has gained popularity in recent years for measurement of methane emission from wetlands, has
typically high temporal resolution due to its continuous operation andminimal disturbance to the ecosystem
measured. This has led to increasing amount of continuous long-term methane flux measurements from
wetlands and other ecosystems.
The methane flux measurements with higher temporal resolution have revealed partly contrasting results on
the environmental controls of methane emission from different study sites. However, the exponential depen-
dence of methane emission on peat temperature has been well documented (Jackowicz-Korczyński et al.,
2010; Kim, Verma, & Billesbach, 1998; Marushchak et al., 2016; Mikhaylov et al., 2015; Rinne et al., 2007).
In part of the measurements, no simple dependence of methane flux on water table has been observed
(Brown et al., 2014; Jackowicz-Korczyński et al., 2010; Rinne et al., 2007), while in others the higher methane
emissions are observed during higher water tables (Mikhaylov et al., 2015). The contrasting results may be
due to differences in the site characteristics and counteracting processes. As a large part of the methane
emission in sedge dominated wetlands is transported via diffusion in the aerenchyma, it bypasses the
potential oxidation in the aerobic acrotelm. This would lead to relative insensitivity of methane emission to
water table position (WT). However, at very low WTs the roots of aerenchymatous plants do not reach the
anaerobic layers inhibiting the plant mediated transport. Conversely, very high water table can submerge
the above-ground parts of aerenchymatous plants, leading also to lower methane emission.
At some of the measurement sites the observed methane emission has a clear diel cycle, with higher diurnal
emission (Kim, Verma, Billesbach, & Clement, 1998; Kowalska et al., 2013), while at other sites no such cycle is
observed (Jackowicz-Korczyński et al., 2010; Rinne et al., 2007). The vegetation at sites with strong diel cycle
of methane emission is commonly dominated by species with convective throughflow via root system, such
as Phragmites australis (Brix et al., 1992).
While methane flux measurements with eddy covariance are now accumulating, very few published time ser-
ies of methane fluxes measured by eddy covariance technique spanning over 5 years exist. Knox et al. (2016)
have presented a 6.5-year time series measured at a Californian rice paddy, and Li et al. (2016) used 5 years of
data from Lompolojänkkä, Finland, and 6 years from Siikaneva, Finland, to evaluate methane model perfor-
mance. Also, many ecosystem scale studies on methane emission have not combined carbon dioxide fluxes
in the analysis. Therefore, our knowledge on interannual variations in ecosystem scale methane emissions is
limited and the relations with environmental factors have been derived from seasonal scale measurements.
This might partly explain the contrasting results on environmental controls because in seasonal time scales
abiotic and biotic factors that regulate methane emissions, such as temperature, water level, leaf area, and
substrate availability via plant production or net ecosystem exchange (NEE), are strongly intercorrelated.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1088
To reveal the dependencies, a longer period of measurements with a large combination of factors is needed.
Recently, homogeneity of slope approach (HOS) has been used to estimate how much of the interannual
variability of carbon dioxide fluxes can be explained by interannual variability of environmental variables
(Hui et al., 2003; McVeigh et al., 2014; Teklemariam et al., 2010). This approach has not thus far been used
for methane fluxes.
As the greenhouse gas balances, including methane emission, of northern peatlands can change either natu-
rally (Mathijssen et al., 2016, 2017) or due to human intervention (Mathijssen et al., 2017; Petrescu et al., 2015),
it is crucial to understand the dynamic responses of methane emission to abiotic and biotic drivers. Also,
northern latitudes are expected to experience faster than average climate warming (IPCC, 2013), which
makes this understanding even more important.
Much of previous work on the relations of methane emission and various abiotic and biotic drivers have been
based on spatial variation (Bubier et al., 2005; Riutta et al., 2007; Turetsky et al., 2014; Whiting & Chanton,
1993). Consequently, relations between methane emission, GPP, and WT may be due to the different plant
communities and associated microbial communities at different microtopographic features, rather than
direct water table effect. Thus, the eddy covariance measurements with high temporal resolution and several
years of coverage can be used to assess the validity of temporal relationships drawn from spatial variation,
that is, space-for-time analogy.
We have conducted eddy covariance flux measurements of methane at a boreal wetland ecosystem since
2005, which to our knowledge constitutes at present the longest methane flux time series by eddy covar-
iance. Our aim in this paper is to quantify the annual balances, seasonal cycles, and interannual variability
of methane fluxes from this ecosystem. Furthermore, the aim is to analyze the relationships of the measured
methane fluxes to the abiotic environmental conditions, as well as biotic substrate control measured as car-
bon dioxide fluxes, at different temporal scales. We will also use the HOS approach to estimate how much of
the interannual variation can be explained by interannual variation of environmental variables.
2. Materials and Methods
2.1. Measurement Site
The data analyzed in this study weremeasured at a fen site at Siikaneva wetland complex in Southern Finland
during years 2005–2015. Siikaneva wetland complex comprises 12 km2 of boreal wetland, including minero-
trophic and ombrotrophic areas. The oldest basal dates of the wetland complex are over 10,000 years, and
the wetland has undergone considerable lateral expansion even within the last 5,000 years. Deepest peat
layers of the wetland complex are over 6 m (Mathijssen et al., 2016). The climate is boreal with no permafrost,
with annual average temperature of 3.3°C and annual rainfall of 710 mm during the period 1971–2000 (Drebs
et al., 2002). The spring thaw typically occurs at the fen in April–May. The fen gets snow cover in November–
December and freezes by the end on December. Thus, the snow and ice clad period is January–March and
snow-free period June–October.
The flux measurements analyzed in this study were conducted in the minerotrophic fen part of the wetland
(61°500N, 24°120E, 162 m above sea level). The vegetation at the site is dominated by peat mosses (Sphagnum
balticum [Russow] Russowex C.E.O. Jensen, Sphagnum majus [Russow] C.E.O. Jensen, and Sphagnum papillo-
sum Lindb.), Sedges (Carex rostrata Stokes, Carex limosa L., and Eriophorum vaginatum L.), and Rannoch rush
(Scheuchzeria palustris L.). The measurements in this site began in 2005. The first results of eddy covariance
flux measurements of carbon dioxide and methane have been published by Aurela et al. (2007) and Rinne
et al. (2007), respectively. In addition, manual chamber flux measurements of carbon dioxide and methane,
conducted in 2004 and 2005, as well as vegetation characteristic are described by Riutta et al. (2007).
Standing plant biomass of the site is presented by Laine et al. (2012). The measurement site is a class 2
ecosystem site in the ICOS research infrastructure (https://www.icos-ri.eu/) since November 2017.
2.2. Flux Measurements
The measurements of carbon dioxide and methane exchange have been conducted nearly continuously
since 2005 using the eddy covariance technique (e.g., Aubinet et al., 2012). Over the 11-year measurement
period, there have been some longer data gaps and changes in the measurement setup, due to development
of technology, breakage of instruments, and logistical demands. The instruments used for eddy covariance
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1089
measurements are listed in Table 1. There have been several changes of methane analyzer and a small
change (7 m) in the location of the measurement tower in February 2010.
The different methane analyzers used at the site have been previously compared by Peltola et al. (2013,
2014). According to their results there are practically no systematic differences between the analyzers used
in this study, the differences being about 4% for the seasonal methane emission. Also, the two individual
sonic anemometers used were compared to each other at the site, and they are unlikely to cause a major
(>10%) systematic difference in gas fluxes.
The postprocessing of the eddy covariance flux data has been done with EddyUH postprocessing software
(Mammarella et al., 2016). The fluxes were calculated as half-hourly averages using block averaging, sector-
wise planar fitting, and empirical high-frequency correction according to Mammarella et al. (2009).
2.3. Ancillary Measurements
The ancillary measurements conducted continuously at Siikaneva include air temperature (Tair) and relative
humidity (Rh), surface moss temperature (Tm), peat temperatures at several depths between 5 and 50 cm
in the peat profile (T5, T20 T35, T50), photosynthetic photon flux density (PPFD), net radiation, precipitation,
and WT. The zero level of the WT must be taken as indicative only, as we do not have detailed description
of the microtopography of the fen surface, needed to tie the water table measurement to surface statistics.
We also used temperature and precipitation data from nearest weather station (Juupajoki Hyytiälä) of Finnish
Meteorological Institute, located 5 km east from the Siikaneva measurement site. Also, global radiation data
from Hyytiälä were used to fill some gaps in radiation data.
2.4. Gapfilling of Carbon Dioxide Fluxes
To obtain daily, monthly, and annual values of NEE of carbon dioxide, and its subcomponents gross primary
production (GPP) and ecosystem respiration (Reco), previously developed and tested gap-filling methods for
wetland fluxes (Aurela et al., 2001, 2007) have been applied. Prior to gapfilling, the data were filtered to
remove data with stationarity test value over 1 (Foken &Wichura, 1996). Friction velocity threshold was deter-
mined based on change-point detection, similarly as in Barr et al. (2013). This analysis gave values between
0.15 and 0.17 m s
1 depending on which periods were used in the analysis. In general, we used 0.17 m s
1 as
the threshold. Also, data with wind direction from pump house and instrument shelter (330°–30°) and periods
when the instrumentation was malfunctioning were removed prior to analysis.
In general, the gap-filling algorithm depends on PPFD, air temperature, andWT. These parameters were mea-
sured on site. In the case of shorter data gaps in PPFD, temperature or WT missing values were gapfilled with
linear interpolation. In the case of gaps longer than a couple of hours, data from Hyytiälä station (Hari &
Kulmala, 2005), located 5 km to the east, were used for PPFD and temperature, whereas data from awater table
measurement at the same peatland complex, about 1 km northwest (for location, see Mathijssen et al., 2016),
Table 1
Instruments Used in the Flux Measurement System
Period Sonic CH4 analyzer CO2 and H2O anal.
2005–2007 Metek USA-1 (a) Campbell TGA-100 Li-Cor 7000
2008–2009 Metek USA-1 (a) Los Gatos RMT-200 Li-Cor 7000
2010 Metek USA-1 (a) Los Gatos RMT-200 Li-Cor 7000
Picarro G1301-f
2011 Metek USA-1 (a) Picarro G1301-f Li-Cor 7000
2012 to January 2014 Metek USA-1 (b) Los Gatos RMT-200 Li-Cor 7000
2014/2 to November 2014 Metek USA-1 (b) Los Gatos GGA-24EP Li-Cor 7000
Los Gatos GGA-24EP
December 2014 to April 2015 Metek USA-1 (b) Los Gatos RMT-200 Li-Cor 7000
Los Gatos GGA-24EP
April 2015 to September 2015 Metek USA-1 (b) Los Gatos GGA-24EP Li-Cor 7000
Los Gatos GGA-24EP
Note. Instruments used in this study are indicated in bold.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1090
were used for water table data. During years 2010 and 2012 PPFD data were missing, and data from Hyytiälä
station were used.
The equation used to gapfill the carbon dioxide flux time series can be written as
NEE ¼ PI α PPFD GPmax
α PPFDþ GPmax þ R0 1þ
b1
1þ exp WTþ b2ð Þ=b3½ 

 
 exp E 1
T0

 1
Tair þ T1
 
 
; (1)
where NEE is the net ecosystem CO2 exchange, GPmax is the gross photosynthesis rate in optimal light con-
ditions (mg m
2 s
1), PI is an empirically determined effective phytomass index (Aurela et al., 2001), PPFD is
the photosynthetic photon flux density (μmol m
2 s
1), α is the initial slope of GPP versus PPFD (mg μmol
1),
R0 is the rate of Reco at 10°C (mg m

2 s
1), WT is water table position (cm), E is a physiological parameter (in
degree Kelvin), Tair is the air temperature, T0 = 56.02 K, and T1 = 227.13 K (Lloyd & Taylor, 1994). The daily PI
value was calculated by subtracting the nocturnal (PPFD< 20 μmol m
2 s
1) respiration flux from the diurnal
(PPFD >500 μmol m
2 s
1) flux as a 7-day running mean. During summertime (approximately from April till
mid-October), E was determined by fitting the temperature to the nocturnal flux data through the year. The
other model parameters b1, b2, and b3 are fitted only once, while α, GPmax, and R0 are fitted in biweekly
periods: An R0 value was first determined by fitting the respiration (second term in the right-hand side of
equation (1)) to the nocturnal NEE for each of these periods, then the values of α and GPmax were obtained
by fitting the NEE equations to the diurnal and nocturnal NEE. During winter when no uptake of CO2 was
observed, the gaps were filled by a running mean of 2 days. The partitioning of NEE to GPP and Reco was
conducted also with equation (1).
2.5. Methane Flux Analysis
The aim of our analysis was to reveal the dependencies of methane fluxes on possible driving variables at
different temporal scales. These include air and peat temperature, humidity, solar radiation, WT, and GPP.
Methane fluxes at the site have earlier been shown to have no systematic diel cycle (Rinne et al., 2007). We
tested this with this much larger data set. First, we calculated the mean and median diel cycles for each
month using the whole data set. Second, we analyzed the environmental controls of hourly variability of
the methane fluxes using generalized linear models (GLMs) by stepwise regression. To analyze the variation
in time scales shorter than a day, we removed the longer-term variation by subtracting corresponding 24-hr
window running-mean-average from each flux value, prior to this statistical analysis.
For analysis of seasonal and interannual variations of methane flux, we averaged the half-hourly values to
daily means. Monthly mean methane fluxes were calculated using daily averages. Before averaging, the data
were filtered by the above-mentioned criteria. All the daily mean methane fluxes comprising of less than 16
half-hourly values were rejected from further analysis. Similarly, themonthly mean value was rejected if it was
composed of less than 10 daily values. After these criteria 53,381 half-hourly (55%), 1,572 daily (39%), and 67
monthly (51%) methane flux values remained for the analysis.
The relations of daily mean methane fluxes to various environmental parameters were first investigated by
linear correlations between methane flux and these parameters. As peat temperature has been shown to
exert exponential behavior to methane fluxes (e.g., Jackowicz-Korczyński et al., 2010; Kim, Verma, &
Billesbach, 1998; Mikhaylov et al., 2015; Marushchak et al., 2016; Rinne et al., 2007), logarithm of methane
emission was also included as a variable in correlation matrix.
In the analysis of the connection between GPP and methane emission, we evaluated the correlations at daily
andmonthly time scales. For the analysis of time lag betweenmethane emission and GPPwe calculated cross
correlations between daily averages of fluxes with different time lags between 0 and 90 days. This analysis
proceeded as follows: (1) We calculated correlation coefficient between time series by aligning the two data
columns, nongapfilled daily averagemethane emission, and GPP, with no time lag applied. Before calculating
the time lag, we removed rows in which either of the data was missing. (2) We took the original data columns,
applied 1-day lag, that is, moved one column one step relative to the other. Again, before calculating the time
lag, we removed rows in which either of the data was missing. (3) We repeated step 2, with applying 2-day
lag. This procedure was repeated over a 90-day period, yielding a correlation coefficient for each time lag.
We also performed the same analysis for temperature-normalized methane emission and GPP. For this, we
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1091
normalized the daily average methane emissions as described in section 2.5.1. Only summertime data (June–
September) were used for the analysis of normalized methane emission due to high variability of normalized
emission in wintertime.
We analyzed the interannual variability of methane flux and its relation to environmental parameters by two
approaches. First, we correlated summer season averages from June to September of different years. In the
analysis of June–September averages, only those years with monthly averages available for each of these
4 months were used (N = 5–6). We did not use any gapfilling of methane fluxes in these analyses to avoid
possible self-correlations. This could occur if interannual variation of methane fluxes, gapfilled, for example,
using temperature, was related to interannual variation of temperature. Second, we used HOS approach (Hui
et al., 2003; see section 2.5.2). The HOS approach has previously been used to study the interannual variability
of carbon dioxide fluxes at peatlands (McVeigh et al., 2014; Teklemariam et al., 2010) but not for methane. For
HOS approach we used nongapfilled daily average methane fluxes.
The annual methane budgets were calculated by gapfilling the methane flux data on daily basis using the
model with best correspondence with the measured data, as quantified by coefficient of determination (r2).
In the following two paragraphs, we describe two approaches for estimating the relations of the measured
daily methane fluxes to environmental parameters. First, we use heuristic approach, based on our prior
knowledge on methane emissions from boreal peatlands. Second, we use a statistical approach, using
GLMs created by stepwise regression.
2.5.1. Heuristic Model
In principle, an algorithm describing the dependence of methane emission on multiple environmental para-
meters can be written as
F ¼∏i f i xið Þ; (2)
in which the flux, F, is a product of independent algorithms, fi, for each environmental parameter, xi. This
approach is similar to the one used in emission models for nonmethane hydrocarbons from vegetation
(e.g., Guenther et al., 2012). As methane emission correlated best with peat temperature at 35 cm depth,
and as it has previously been shown to have an exponential relation with temperature, we fitted first an expo-
nential function,
f 1 Tð Þ ¼ a1ea2T ; (3)
to the observed daily average fluxes. Here T is the peat temperature at chosen depth and a1 and a2 are
empirical coefficients. The depth off the temperature measurement was chosen based on best correlation
between temperature and methane flux. The fitting was done using nlinfit function of Matlab (R2015b).
Gross primary production can drive methane production via root exudates. Thus, we can use this as second
parameter in our heuristic model. In order to remove the dominant effect of temperature on emission, we
first normalized the daily average methane emission by dividing it with the temperature dependence
function (equation (3)),
Fn ¼ Ff 1 Tð Þ : (4)
The intersite comparison by Whiting and Chanton (1992) showed linear relation between methane emission
and NEP. Hence, we analyzed the relation between methane emission normalized by temperature (equa-
tion (4)) and GPP using a linear relation, in the form of
f 2 GPPð Þ ¼ c1 þ c2GPP; (5)
where c1 and c2 are empirical coefficients.
As the WT is often referred to be of major importance to methane emission, we also investigated the relation
of methane emission and WT. We investigated this relation both by normalizing the daily average methane
emission first by temperature dependence (equation (4)) and second by normalizing with both temperature
and GPP dependence.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1092
Previous studies have reported, for example, methane fluxes decreasing exponentially with descending WT
(Bubier et al., 2005) and unimodal relations (Turetsky et al., 2014). To better visualize the shape of the water
table relation we bin-averaged the normalized daily averagemethane flux data according toWT to 5-cm bins.
In order to avoid the effects of ice and snow cover, we used only data from June to September to derive these
relations. These months have highest vascular LAI (Riutta et al., 2007) and also the highest methane fluxes.
There are likely to be various processes leading to both decrease and increase in the methane emission from
fens with low WT (Raivonen et al., 2017). Review by Turetsky et al. (2014) also shows evidence on unimodal
water table dependence, at least from certain types of wetlands. To mathematically describe the effects of
the counteracting processes on methane fluxes, we can use two sigmoid curves, representing increase and
decrease of methane emission with water table. Combining two sigmoid curves with opposite tendencies
leads to an equation with expected behavior,
f 3 WTð Þ ¼ b11þ eb2 WTþb3ð Þð Þ 1þ e
b4 WTþb5ð Þð Þ ; (6)
Figure 1. Seasonal and diel cycles of median (a) methane and (b) carbon dioxide fluxes, based on nongapfilled data.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1093
where WT is the water table position and b1–b5 are empirical coefficients. In this formulation, the coefficient
b1 defines the level the sigmoid curve reaches, coefficients b2 and b4 defines the slope of the rising or
descending part of the curve, and coefficients b3 and b5 the location of the rising and descending parts
along WT axis. The formulation approaches zero for high and low water tables and can have asymmetric
slopes across maximum of the curve, depending on coefficients b2 and b4. We fitted equation (6) to bin-
averaged normalized methane fluxes using nlinfit function of Matlab (R2015b).
Finally, daily methane fluxes were calculated using a model comprising either one, two, or three parameters.
For example, the equation comprising all three parameters is written as
F ¼ f 1 T35ð Þf 2 GPPð Þf 3 WTð Þ; (7)
where f1 is equation (3), f2 is equation (5), and f3 is equation (6). The calculated emissions were compared to
observations by goodness of fit.
2.5.2. Stepwise Regression and Generalized Linear Models
Generalized linear models describe response variable with a linear combination of functions depending on
input parameters. In the case of simple model with linear responses and no interactions between variables,
this can be written as
Y ¼ β0 þ
X
i
βiXi; (8)
where Y is the response vector, in the case of this analysis methane flux, and Xi a column of input
matrix, that is, environmental parameter. In more general case we can have interactions between
Figure 2. Time series of (a) daily averages of methane flux (nmol m
2 s
1), (b) gapfilled daily net ecosystem exchange
(μmol m
2 s
1), (c) daily average air temperature (gray) and peat temperature at depth of 35 cm (black; °C), and (d)
daily average water table position (cm).
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1094
parameters (mathematically their products) and even interaction with
categorical parameters (i.e., different slopes for observations with different
categorical parameter).
We created the model using first stepwise regression (Matlab stepwi-
seglm) using daily average data for most parameters and daily gapfilled
sums for carbon dioxide fluxes. We selected the variables to include in
the stepwise regression by examining the correlation diagram. Based on
correlation coefficients we first selected the parameter with highest corre-
lation. We then excluded those that are highly correlated with the para-
meter selected first. After final selection of the model parameters by
stepwise regression, we refitted the model using linear regression
(Matlab fitglm). In order to examine seasonal differences in drivers of
methane emission, we conducted the analysis also separately for snow-
free season (June–October) and snow period (January–March).
2.5.3. Homogeneity of Slope Approach
We also use the GLM as a tool to investigate the interannual variability, fol-
lowing the approach of HOS model (Hui et al., 2003). In this approach, we
used stepwise regression to create a model without year as category vari-
able, that is, fixed slope model (section 2.5.2). As second step, we create a
model using the explanatory variables of the fixed slope model but includ-
ing year as categorical variable with linear regression (matlab fitglm),
leading to variable slopes model. As in 2.5.2, we include interactions
between parameters and allow quadratic relations. If the variable slope
model is significantly better than the fixed slope model, according to F
test, we interpret this difference as indicating unresolved interannual
variability or “functional change” (Hui et al., 2003).
After creating and testing the models, we calculated the components of
sum of squares of total deviation to quantify the sources of variation in
methane fluxes. The total sum of squares difference (SStot) between the
variable slope GLM and observations can be divided into its compo-
nents as
SStot ¼ SSsc þ SSic þ SSf þ SSe; (9)
where SSsc is due to seasonal variation of drivers, SSic is due to interannual
variation of drivers, SSf is unresolved interannual variation (functional
change), and SSe residual error. The formulations of these components
follow those by Hui et al. (2003).
2.5.4. Gapfilling the Methane Flux Data
The gapfilling method was selected by the performance of heuristic mod-
els and GLMs. As carbon dioxide fluxes are likely to suffer partly from
similar data gaps as methane fluxes, we used a hierarchy of gapfilling pro-
cedures. Accordingly, the first step of gapfilling was conducted using the
model that performed best against the data. The gaps remaining after this
step were filled with model that did not use, for example, carbon fluxes as
input. Finally, the remaining gaps were filled with interpolation.
3. Results
3.1. Hourly Variation
The half hourly methane fluxes varied from 4.0 to 94 nmol m
2 s
1 (5%
and 95% percentiles). The contour plots of carbon dioxide and methane
fluxes on month-hour plane (Figure 1) show strong seasonal variation
for both carbon dioxide exchange and methane emission, but no diel
Figure 3. Monthly average nongapfilled methane emission against monthly
average gross primary production.
Figure 4. Correlation (r) as a function of time lag applied between time
series. Only correlations with p < 0.01 are shown.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1095
cycle for methane flux. The GLMs created by stepwise regression for
deviation of flux from its daily running mean had low explanatory
power, with r2 < 0.1. The full methane flux data set showed a minor
diel cycle (Figure S1 in the supporting information), but removing the
data obtained by Los Gatos RMT200 methane analyzer mostly removed
the diel cycle (Figure 1).
3.2. Seasonal Variation
The daily averaged methane fluxes and daily average gapfilled NEE
showed pronounced seasonal cycles (Figure 2), which led to significant
correlations between fluxes and practically all environmental parameters
(Figures 2 and S2). Methane flux correlated best with peat temperature
at 35 cm depth (r = 0.82), temperature at 20 cm depth (r = 0.81), and
GPP (r = 0.80). The logarithm of methane flux correlated best with tem-
perature at 35 cm depth (r = 0.90), peat temperatures at other depths
(r = 0.85–0.89), and GPP (r = 0.79). The correlation between the logarithm
of methane emission and peat temperature at 35 cm depth was signifi-
cantly higher than that between 5 and 50 cm (Fisher’s r to z transf.,
p < 0.01) but did not differ significantly from that between 20 cm.
However, for all individual years except 2011, the best correlation was
found for peat temperature at 35 cm depth. The average seasonal cycle
of methane emission showed maximum emissions in July and August
(average 73 nmol m
2 s
1), with small but positive fluxes also during win-
ter (January–March average 6.7 nmol m
2 s
1). While air temperature was
typically well below zero in winter, and the surface of the peatland freezes
to the depth of about 20 cm, the peat temperature at the depth of 35 cm
stayed positive throughout the winter. The lowest WTs were observed in
summer, but the interannual variation was large. The average water table
during the measurements was 
5 cm, and the average during ice- and
snow-free June–September period was 
13 cm, referenced to a some-
what arbitrary reference point (see section 2.3). During some years but
not all, a small methane burst connected to thaw period in spring was
observed (Figure S2).
Gross primary production had a high correlation with methane flux
(Figure S2). By averaging the data to monthly averages, we observe a
hysteresis-like behavior (Figure 3), with methane emission lagging behind
GPP. The lag-time analysis (Figure 4) showed the correlation between methane emission and GPP to have a
broad peak at lag times of 5–20 days. However, methane flux normalized by its temperature dependence did
not show similar time lag (Figure 4).
3.3. Heuristic Model
The daily average methane emission against peat temperature at the depth of 35 cm showed exponential
behavior (Figure 5a). Equation (3), fitted to the data, described the methane fluxes with r2 = 0.73
(Figure 7a). The apparent Q10 obtained by fitting to the whole data set was 5.3, with annual values ranging
from 3.8 to 12. The methane emission normalized by equation (4) showed positive correlation with GPP
(Figure 5b; r = 0.53, p < 0.01). There was practically no correlation between WT and methane flux.
However, bin-averaged data suggest a nonmonotonous relation to WT (Figure 6a). Especially, after normal-
izing the methane emission with temperature (equation (3)) and GPP (equation (5)), we observe mostly no
dependence on water table, except for the very lowest and highest WTs (Figures 6b and 7).
The inclusion of GPP by linear equation (equation (5)) improved the correspondence of model to data
(r2 = 0.80), but inclusion of WT (equation (6)) did not improve the correspondence, with r2 = 0.74 for T-WT
algorithm and r2 = 0.80 for T-GPP-WT algorithm.
Figure 5. (a) Daily average methane fluxes against daily average peat tem-
perature at the depth of 35 cm. The black line indicates the equation (3)
fitted to the data (a1 = 0.00682 μmol m

2 s
1, a2 = 0.167°C

1, r2 = 0.74). (b)
Daily average temperature-normalized methane flux against gross primary
production. The black line indicates linear fit to data (equation (6), c1 = 0.585,
c2 = 0.162 [μmol

1 m2 s], r2 = 0.28).
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1096
3.4. Generalized Linear Models
The GLMs in annual basis were created by stepwise regression using the
parameters shown in the Table 2. The first parameter that was selected
by the stepwise regression algorithm was peat temperature, and the sec-
ond one was GPP. The final fixed slopes model covering full annual cycles
can be written as
ln F ¼ β0 þ β1T þ β2GPPþ β3T GPPþ β4T2 þ β5GPP2; (10)
with coefficients in Table 3.
The variable slope model for full annual cycles, created using the same
explanatory parameters as the fixed slopes model but with year added
as categorical variable, can be written as
ln F ¼ β0 þ β1T þ β2GPPþ β3T GPPþ β4T2 þ β5GPP2
þ
X
yr
β6;yr þ β7;yrT yr þ β8;yrGPPyr
 
: (11)
Thus, both models include peat temperature (35 cm) and GPP as drivers.
These models correspond to the observed fluxes well, with r2 = 0.84 for
fixed slope model and r2 = 0.89 for variable slope model (Figure 8). The
variable slope model was significantly better according to F test (p< 0.01).
As the snow-free period (June–October) methane fluxes correlated best
with the temperature at 20 cm, we used this for GLM for summer. For
the snow-free period the stepwise regression selected peat temperature
as the first explanatory variable and GPP the second. As for the whole year,
the inclusion of year as category variable leads to variable slope model
that was significantly better than the fixed slope model. Although the
parameters in both models were the same for the summer time and whole
year, the formulations were slightly different.
Wintertime (January–March) methane emissions were very small. Fixed
slope GLM created with stepwise regression included peat temperature
as parameter, but the explanatory power of the model was very low
(r2 = 0.16). The variable slope GLM had only year as parameter and was
significantly better with r2 = 0.42.
3.5. Gapfilling and Annual Fluxes
Based on the performance of the heuristic model and GLM, we chose to
gapfill the methane fluxes using mainly GLMs. As the best performing
model was the variable slope model with peat temperature, GPP, and year as variables, we used this as first
step. As second step, we used a variable slope model that did not employ GPP but used other variables (air
temperature, global radiation, and peat temperature). Third step did not use global radiation as this para-
meter was missing during one period. Finally, remaining gaps were filled by linear interpolation. The mea-
sured daily average data covered 48% of the period 2005–2014. First step increased data coverage to 86%,
second to 90%, and third to 94%, with remaining 6% filled up with interpolation. The gapfilled methane flux
time series is shown in Figure 9. The average annual methane emission with standard deviation was
9.7 ± 2.5 gC m
2. 88 ± 3% of the annual emission occurred in May–November, which are mostly snow and
ice free. 6 ± 2% occurred in snow and icebound January–March. The annual and monthly values are listed
in Tables 4 and S3 in the supporting information. Annually, the methane emission was 2.7 ± 0.4%
(mean ± standard deviation) of GPP. Over 10 years, methane emission was 18% of the NEE.
3.6. Interannual Variation
The annual cumulative methane flux varied from 6.0 to 14 gC m
2, while the mean annual NEE, GPP, and Reco
were 
54 gC m
2 (
96 to 27 gC m
2), 390 gC m
2 (260–460 gC m
2), and 330 gC m
2 (290–390 gC m
2),
Figure 6. (a) Daily average temperature-normalized methane flux normal-
ized, together with bin-averages, against water table position. The black
line indicates equation (5) fitted to bin-averages (b1 = 1.07; b2 = 0.388 cm

1;
b3 = 
3.83 cm; b4 = 0.213 cm
1; b5 = 36.3 cm, r2 = 0.06). (b) Daily average
temperature-gross primary production-normalized methane flux normal-
ized, together with bin-averages, against water table position. The black line
indicates equation (5) fitted to bin-averages (b1 = 1.03; b2 = 0.666 cm

1;
b3 = 
3.10 cm; b4 = 0.168 cm
1; b5 = 38.0 cm, r2 = 0.06).
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1097
respectively. The relative variations of these fluxes, expressed as coefficient of variation (standard deviation of
the annual flux, divided by the mean of annual fluxes), were 0.25, 0.73, 0.17, and 0.13 for methane emission,
NEE, GPP, and Reco, respectively.
The interannual variation of seasonal nongapfilled June–September
methane emission showed significant correlation with only GPP (r2 =
0.78, p = 0.046, Figure 10). Correlations with other parameters (peat
temperature, WT, and NEE) were not significant.
The HOS analysis showed that the variable slope model performed signifi-
cantly better than fixed slope model (F test, p < 0.01) for both the whole
annual cycle and for summertime only. On the annual scale, the seasonal
variation (SSsc) explained 79% of the sum of squares variation, the inter-
annual variation due to variation in explaining factors (SSic) 7.0%, interann-
ual variation not accounted for by explaining factors (functional change,
SSf) 5.3%, and random variation (SSe) 9.1%. Thus, 43% of the interannual
variability was due to functional change.
4. Discussion
4.1. Temporal Variation of Methane Fluxes
The measured methane fluxes showed pronounced annual cycles over the
measurement period, but no diel cycle. The diel cycle shown by the whole
data set (Figure S1) is likely to be mostly an artifact created with imperfect
Figure 7. Performance of heuristic models against data. (a) Temperature algorithm. (b) Temperature-water table algorithm.
(c) Temperature-gross primary production (GPP) algorithm. (d) Temperature-GPP-water table algorithm.
Table 2
Parameters Used in Stepwise Regression to Generate the Generalized Linear
Model
Model Parameters
Whole period Air temperature
Vapor pressure deficit
Global radiation
Peat temperature at 35 cm
Water table position
Gross primary production (GPP)
Snow-free period Air temperature
Vapor pressure deficit
Global radiation
Peat temperature at 20 cm
Water table position
GPP
Winter Air temperature
Vapor pressure deficit
Global radiation
Peat temperature at 35 cm
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1098
density correction to Los Gatos RMT200 data by WPL approach (Webb et al., 1980). As Campbell TGA-100
dried the sample airstream, and Picarro G1301-f and Los Gatos GGA-24EP measured water vapor in their
optical cell, enabling point-by-point density calculation, using the data from these analyzers leads to no
systematic diel cycle (Figure 1). Our analysis with stepwise regression did not show any clear relation of
methane fluxes at hourly time scales and any environmental parameters. The sedge dominated fens have
different emission dynamics than Phragmites dominated wetlands (Kim, Verma, Billesbach, & Clement,
1998; Kowalska et al., 2013). The apparent diel cycle of the data measured with LGR RMT200, probably
due to water vapor related density and line broadening effects, is a relatively minor one. As the quality
control routines employed in eddy covariance data processing are more likely to remove nocturnal
measurements, the daily emission derived using simple daily averages may be an overestimate. However,
calculating the monthly average emission from medians in Figure S1 using 8 afternoon hours leads to
overestimations that are below 10% for all months except May, when the overestimation was 12%. Carbon
dioxide fluxes observed here showed diel and annual cycles characteristic to boreal ecosystems (e.g.,
Aurela et al., 2007; Suni et al., 2003), with uptake during summer days and release to the atmosphere
during night and winter (Figure 1b).
The spring pulse, when observed, was small in relation to the summertime fluxes, similar to the observations
previously reported at the same site by Rinne et al. (2007) and at a boreal peatland in Komi Republic of
Russian Federation by Mikhaylov et al. (2015). This modest magnitude contradicted with Hargreaves et al.
(2001), who observed a significant fraction of annual methane emission from subarctic Kaamanen mire to
be emitted during spring pulse. The importance of spring pulse may indeed be higher at more northern sites,
such as Kaamanen, with deep frozen layer in winter. In south-boreal Siikaneva, with only about 20-cm frozen
layer, much of the methane produced in wintertime may escape through ice and snow, as the frozen layer in
peatlands is more porous than, for example, lake ice, containing in many places frozen plant parts, such as
Table 3
Coefficients for Fixed Slope and Variable Slope Generalized Linear Model for the Whole Measurement Period
Fixed slope model, equation (10) Variable slope model, equation (11)
Coefficient p-value Coefficient p-value
β0 
5.2286 β0 
4.9533
β1 0.10721 9.3 × 10

13 β1 0.086069 5.3 × 10

7
β2 0.36583 9.5 × 10

10 β2 0.35254 9.8 × 10

9
β3 
0.032327 3.0 × 10
5 β3 
0.021952 0.0034
β4 0.0044326 0.00095 β4 0.0033769 0.016
β5 0.036195 0.0021 β5 0.021844 0.035
β6,2006 
0.26346 0.00062
β6,2007 
0.38856 3.0 × 10
8
β6,2008 
0.27619 0.019
β6,2010 
0.079762 0.45
β6,2011 0.03737 0.68
β6,2012 
1.2885 6.3 × 10
38
β6,2014 0.38364 7.6 × 10

6
β7,2006 
0.039734 0.059
β7,2007 
0.0024854 0.88
β7,2008 0.05356 0.00026
β7,2010 0.0094902 0.57
β7,2011 
0.015824 0.27
β7,2012 0.11631 2.8 × 10

8
β7,2014 0.0022082 0.89
β8,2006 0.27084 0.0025
β8,2007 0.059658 0.39
β8,2008 
0.15737 0.0020
β8,2010 0.018061 0.73
β8,2011 
0.018087 0.73
β8,2012 
0.04172 0.54
β8,2014 
0.12246 0.032
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1099
moss and sedges. Still, emission during the summertime clearly domi-
nated the annual emission, with emission of snow and ice clad period
being of minor importance.
4.2. Relation of Methane Flux to Environmental Parameters
Daily average methane fluxes correlated significantly with air temperature
and peat temperature measured at different depths, as well as other abio-
tic parameters (Figure S2). These parameters also show highly significant
correlations with each other, indicating largely coherent seasonal cycles.
The highest correlations were observed between logarithm of methane
flux and peat temperatures at 20–35 cm depth, similarly to the earlier
observation at the site by Rinne et al. (2007). This is in line with the
common observation of relation of methane emission on temperature
being exponential (e.g., Jackowicz-Korczyński et al., 2010; Kim, Verma, &
Billesbach, 1998; Kim, Verma, Billesbach, & Clement, 1998; Marushchak
et al., 2016; Mikhaylov et al., 2015). The temperatures at different depths
of peat layer have effects on different processes leading to net methane
emission. For example, temperature responses of methane production
and consumption differ from each other (Dunfield et al., 1993) which
can lead to confounding effects on observed temperature relations.
Consequently, the statistical response of measured net methane emission
to temperature does only quantify link empirically, but does not disentan-
gle the underlying mechanisms.
In the previous study using data from 2005 at the site by Rinne et al. (2007),
exponential relation between methane emission and peat temperature
was observed at temperatures below 12°C, with no correlation above
12°C. By using the current 11-year data set, the exponential relation
between peat temperature and methane flux was confirmed for the whole
temperature range, with no upper limit (Figure 5). The reason for the lack
of temperature relation above 12° in year 2005 may be related to rapid
ascent of water table at the onset of peat temperatures above 12°C.
The values of Q10 presented in results were calculated from seasonal time
series and are affected by confounding factors, such as phenology. Thus,
we refer to them as apparent Q10. While part of the apparent Q10 is likely
to be due to the confounding factors, the temperature is also likely to have
a direct effect on the metabolic activity of methane producing Archaea.
The published apparent Q10 values of methane fluxes, measured by eddy
covariance, range from 2.4 to 6.6 (Table 5). The apparent Q10 of the whole
data set presented here, 5.3, is within the range of these, although the
interannual variation is large. We should note that different studies have
correlated the methane fluxes with temperature measured at different
depths, which affects the apparent Q10 values. For example, with our data
the apparent Q10 using peat temperature at 5-cm depth would have been
3.1. Thus, difference between Q10 obtained from different sites may reflect
the differences of peat temperature measurements. However, the inter-
annual or seasonal variations in the Q10 may be due to variations in other
controlling factors.
Water table has been commonly indicated to be one of the major controls
of methane emission (e.g., Bubier et al., 2005; Turetsky et al., 2014).
However, in some studies on temporal behavior of ecosystem scale
methane emission, no clear dependence of methane fluxes on WT has
been observed (e.g., Jackowicz-Korczyński et al., 2010; Rinne et al., 2007).
Figure 8. Performance of generalized linear models against data. (a) Fixed
slope model. (b) Variable slope model.
Figure 9. Time series of gapfilled daily average methane flux.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1100
Our analysis also shows no clear dependence of methane emission on
water table, but the bin averaged data suggest a broad maximum of
methane fluxes at intermediate WTs between about 
25 cm and 
5 cm,
with a slight decrease of emission both with low and high water tables
(Figure 6). As water table was between these values for most of the
snow-free period, the water table did not have a major effect on temporal
variation of methane emissions. These water table limits must be taken as
indicative only, as the zero level of water table is not quantitatively tied to
surface microtopography statistics. Similar unimodal behavior of methane
emission with water table was obtained by annual GLM without year as
category parameter. The observed water table dependence is qualitatively
similar to that of NEE as observed by Aurela et al. (2007).
There are likely to be various processes leading to decrease in themethane
emission from fens with low WT. For example, as WT descends, a consider-
able part of roots of aerenchymatous plants may not reach the anaerobic
water saturated peat, thus inhibiting this transport route to the atmo-
sphere. According to Saarinen (1996) majority of roots of many vascular
plants in peatlands are in quite shallow layer (0–30 cm), although part of
the root system reaches much deeper. The increased oxidation in the dif-
fusion pathway within nonsaturated peat is likely to increase this effect.
Thus, if the water table were drawn much lower than observed during
the 10-year period reported here, the methane emission would be inhib-
ited more, and the peatland could even turn to a small sink of methane
at extremely low WT.
On the other hand, there can be at least two mechanisms that can lead to
lower methane emissions with very high water table. First, increasing
water table may lead to total submersion of aerenchymatous plants at hol-
lows and lower lawn surfaces, leading to decreasing fluxes with ascending
water table. The review by Turetsky et al. (2014) point out to low emission
from flooded sites. Second, if the ascend of water table is due to rain at the
site, it may create a layer of water with low dissolved methane concentra-
tion on top of peat water with higher concentration. This may slow down
the diffusion through peat matrix in the topmost peat layers, as the diffu-
sion in water is slower than in air-filled pores.
The relative insensitivity of methane flux to WT within the range observed
in this study may indicate a small contribution of methane oxidation as
control of methane emission from sedge fens. This may be due to abun-
dant aerenchymatous vascular plant cover at this type of peatlands, which
leads to most of the methane to be emitted through the aerenchyma.
Thus, the methane bypasses the aerobic peat and moss layers for the typi-
cal range of WTs.
The fundamental difference of the water table relation derived here, and
those represented by, for example, Bubier et al. (2005), Riutta et al.
[2007], and Turetsky et al. (2014), is that the one presented here concerns
temporal variability on subannual time scale, while those mentioned
above refer mostly to spatial variability. In the latter, the plant community
composition, that is, abundance of aerenchymatic plants and dwarf
shrubs, may play an important role in either promoting or inhibiting the
methane emission (Joabsson et al., 1999; Wang et al., 2015), and thus, it
is not obvious that the space-to-time analogy is applicable to seasonal
time scales.
Even though the methane is strongly correlated with temperature on a
seasonal time scale, their interannual variation does not show significant
Table 4
Annual Cumulative Values of Methane Emission (FCH4), Net Ecosystem
Exchange (NEE), Gross Primary Production (GPP), and Ecosystem Respiration
(Reco)
FCH4 NEE GPP Reco Tave Precip.
gC m
2 Q10 gC m

2 gC m
2 gC m
2 °C mm
2005 9.0 3.8 
29 380 350 4.4 703
2006 6.0 4.4 27 260 290 4.6 653
2007 7.0 5.9 
48 350 290 4.5 700
2008 8.9 5.2 
78 370 300 4.9 933
2009 9.6 5.9 N.A. N.A. N.A. 3.8 502
2010 11 5.8 
58 350 310 2.5 657
2011 8.8 4.5 
60 450 390 4.9 745
2012 11 6.6 
96 410 300 3.2 907
2013 14 12 
85 460 370 4.9 615
2014 12 11 N.A. N.A. N.A. 5.1 579
2015 N.A. 6.1 N.A. N.A. N.A. 5.5 658
Note. Annually derived Q10. Mean annual temperature (Tave) and annual
cumulative precipitation (Precip.) at Juupajoki Hyytiälä weather station.
Figure 10. June–September nongapfilled average methane flux against (a)
air temperature (°C) at Hyytiälä Juupajoki weather station, (b) precipitation
(mm mo
1) at Hyytiälä Juupajoki weather station, (c) peat temperature (°C)
at 35 cm depth, (d) water table position (cm), (e) net ecosystem exchange
(gC m
2 mo
1), and (f) gross primary production (GPP; gC m
2 mo
1).
The regression line (FCH4 = 0.0600 × GPP 
 2.0420 mgC m
2 h
1) is
included in (f), where the correlation is significant.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1101
correlation. This may be partly due to the smaller data set, 5 years, for June–August period. Furthermore, the
maximum interannual variation of peat temperature at 35 cm depth was only 1.2°C.
4.3. Relation Between Methane Emission and Carbon Dioxide Fluxes
As methanogenic Archaea use mostly relatively fresh root exudates as substrate to methane production
(Chanton et al., 1995; Couwenberg & Fritz, 2012; Ström et al., 2003; Whiting & Chanton, 1993), a link between
carbon dioxide uptake and methane emission can be expected (Joabsson & Christensen, 2001). The methane
fluxes indeed showed correlation with GPP, with hysteresis-like behavior (Figure 3). Following leaf area devel-
opment typical to boreal peatlands (Wilson et al., 2007) and high availability of light in the early summer, GPP
increases rapidly, but methane emission increases at much slower rate. In the late summer GPP decreases
rapidly, but methane emission continues at a relatively high rate.
The lag times between the methane emission and GPP, as shown by the cross-correlation analysis (Figure 4),
were on average 15 days with quite broad peak between about 5 and 25 days. However, temperature-
normalized methane emission did not show any correlation lag. Thus, the lag between GPP and methane
emission is likely to be due to lag between the environmental parameters controlling the GPP and those
controlling methane emission, in the seasonal scale. GPP is controlled largely by PPFD and air or surface tem-
perature, while methane emission by the temperature control and substrate availability of archaea metabo-
lism. Incubation experiments show increased methane production with increasing temperature indicating
temperature control (e.g., Juottonen et al., 2008). The significant correlation of temperature-normalized
methane emission with GPP indicates that the fresh substrate input is an important controlling factor of
methane emission. A rapid link between photosynthesis and methane emission has been shown in pulse
labeling experiments, where 14C label appeared onto emitted methane in less than 1 day (King et al., 2002;
King & Reeburgh, 2002). The difference between these studies and the one reported here is that labeling
experiments show when labeled carbon first appears into emitted methane, while our analysis shows the dif-
ferences in the seasonal cycles of methane emission and GPP.
The interannual variation of methane emission was significantly correlated to that of GPP, when analyzed
for the summer season (June–September). This, together with published observations on 14C labeling
(King et al., 2002; King & Reeburgh, 2002), indicates a causal dependence of seasonal methane emission
Table 5
Published Apparent Q10 Values Based on Temporal Variability of Ecosystem Scale Emission
Reference Ecosystem Depth of temperature measurement Q10
This study Boreal fen 35 cm 5.3 (3.8–12)
Kim, Verma, and Billesbach (1998) Temperate midlatitude
marsh
Average of top sediment 0–15 cm,
under 0.4–0.6 m of water
4.4–6.6
Jackowicz-Korczyński et al. (2010) Subarctic mire 3 cm 3.3–5.3
Mikhaylov et al. (2015) Boreal peatland 15 cm 4.3
Marushchak et al. (2016) Subarctic wet tundra 25 cm 2.4
Table 6
Mean Annual Methane Emission (FCH4), Net Ecosystem Exchange (NEE), Gross Primary Production (GPP), and Their Ratios Obtained by This Study, With Values Reported
in Literature
Site Type Cite
FCH4 NEE GPP Reco FCH4/
NEE
FCH4/
GPPgC m
2 gC m
2 gC m
2 gC m
2
Siikaneva fen Fen This study 9.7 
54 380 330 18% 2.7%
Biebrza Valley Disturbed fen Fortuniak et al., 2017 18 
210 n.a. n.a. 8.6% n.a.
Federseemoor Phragmites fen van den Berg et al., 2016 23 
240 850 610 9.4% 2.7%
Forsinard Bog Levy & Gray, 2015 4.3 
110 575 460 3.9% 0.75%
Schechenfilz Forested bog Hommeltenberg et al., 2014 5.3 
62 n.a. n.a. 8.6% n.a.
Bog lake Fen Olson et al., 2013 16 
35 n.a. n.a. 46% n.a.
Samoylov Island Wet polygonal tundra Wille et al., 2008 2.4 
20 n.a. n.a. 12% n.a.
Mer Bleue Bog Roulet et al., 2007 3.7 
40 n.a. n.a. 9.2% n.a.
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1102
on photosynthesis as carbon input. The correlation of temperature-normalized methane flux with GPP
(Figure 5) indicates also that the correlation between methane emission and GPP in the seasonal time scale
is not purely a result of collinearity of GPP and peat temperature.
The ratio of methane emission to GPP was in the same range as obtained by intersite comparison by Whiting
and Chanton (1993). Themethane emission to GPP ratio (FCH4/GPP), if shown to hold for a range of peatlands,
could be a useful constrain for regional and global peatland methane emissions. Unfortunately, few sites
report data to enable comparisons of methane FCH4/GPP ratios in ecosystem scale (Table 6). Themethane flux
at Siikaneva was 18% of the NEE. This is higher than in the other sites listed in Table 6. However, these sites
were mostly ombrotrophic peatlands with one disturbed fen. Thus, the difference of FCH4/NEE is likely to be
due to ecosystem type.
The observed links between methane emission and GPP are crucial for development of process-based
methane emission models. In such models the methane production typically depends on the available car-
bon pool development (Li et al., 2016; Oikawa et al., 2017; Raivonen et al., 2017; Tian et al., 2010; Walter &
Heimann, 2000; Wania et al., 2010; Watts et al., 2013; Zhuang et al., 2004), in line with our findings. Also,
the possibility to remotely observe abiotic and biotic parameters controlling the carbon dioxide exchange
of wetland ecosystems (Aurela et al., 2004; Lees et al., 2018; Rinne et al., 2009) may be linked to methane
emissions from northern peatlands. However, in order to better understand the dynamic links between
carbon dioxide fluxes and methane emission, measurements of seasonal development of substrate and
dissolved methane pools within anoxic peat layers would be crucial.
4.4. Emission Algorithms
The two approaches we used to create emission parameterization for daily average methane fluxes, the
heuristic and stepwise regression, showed similarities in the driving parameters. Both lead to algorithms in
which peat temperature had a strong exponential control on calculated emission, with further control by
GPP. The effect of WT, if included in a model, was very small or nonexistent.
The usefulness of the models with GPP as explanatory variable in gapfilling is somewhat limited. NEE mea-
surements suffer from many of the same limitations as methane flux measurements, commonly reducing
data coverage and requiring further use of gapfilling algorithms to obtain daily GPP values. Thus, in many
cases, gapfilling with temperature and possibly water table algorithm is more feasible.
The GLMs explained the observed variation in methane emission better than heuristic model, especially
when letting the slopes vary between years. The latter may be somewhat more transparent and can be mod-
ified based on knowledge from external studies.
4.5. Interannual Variation
The relative interannual variability of methane emission from the fen ecosystem was slightly higher than
those of GPP and Reco, but much less than that of NEE. As indicated by HOS analysis and significant correla-
tion of June–September methane emission with GPP, a part of this interannual variability is related to inter-
annual variability of GPP. The HOS approach indicated that more than half of the interannual variability was
accounted for by the interannual variation of GLM drivers, peat temperature, and GPP.
The interpretation of the HOS result, as applied here to methane emission, is different than the analyses
earlier conducted for carbon dioxide (Hui et al., 2003; McVeigh et al., 2014; Teklemariam et al., 2010). In those
studies, only climate parameters were used as drivers, whereas for methane, GPP is indicated to be the
second most important driver of methane emission by stepwise regressions. GPP is not a climate variable,
but itself a result of ecosystem functioning. Thus, the interpretation of interannual variability of methane
emission due to functional change of the ecosystem differs from that of the studies for carbon dioxide. We
may interpret this as functional change in the methane production and oxidation processes.
5. Conclusions
A decade-long data set on ecosystem scale methane emission, measured by the eddy covariance technique,
showed strong seasonal cycle but no systematic diel cycle. Daily average methane emission related mostly
to peat temperature and GPP. Water table had no significant effect to methane fluxes for the observed
WT range. This indicates that space-for-time analogy, commonly used upscale spatial chamber data
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1103
temporally, is not necessarily applicable in seasonal time scales. The variation of methane emission in time
scales shorter than 1 day did not show strong relations with any environmental variables.
The observed relations suggest that methane emission is strongly regulated by temperature controlling, for
example, the metabolic rates of methanogenic Archaea, and by carbon input via GPP. The observed close link
between GPP and methane emission is crucial for development of process-based methane emission models.
Methane oxidation does not seem to be an important control of variations in methane emission for the
typical WT range, as water table did not show strong relation with methane emission.
The interannual variation of summertime methane emission correlates with that of GPP. The HOS approach
showed that interannual variation of GPP, temperature, and other meteorological parameters were not able
to fully explain the interannual variability of methane emission. On the annual scale, the seasonal variation
(SSsc) explained 79% of the sum of squares variation, the interannual variation due to variation in explaining
factors (SSic) 7.0%, interannual variation not accounted for by explaining factors (functional change, SSf) 5.3%,
and random variation (SSe) 9.1%. Thus, 43% of the interannual variability was due to functional change.
References
Aselmann, I., & Crutzen, J. P. (1989). Global distribution of natural freshwater wetlands and rice paddies, their net primary productivity,
seasonality and possible methane emissions. Journal of Atmospheric Chemistry, 8(4), 307–358. https://doi.org/10.1007/BF00052709
Aubinet, M., Vesala, T., & Papale, D. (Eds.) (2012). Eddy covariance—A practical guide to measurement and data analysis. Dordrecht: Springer.
978-94-007-2350-4
Aurela, M., Laurila, T., & Tuovinen, J.-P. (2004). The timing of snow melt controls the annual CO2 balance in a subarctic fen.
Geophysical Research Letters, 31, L16119. https://doi.org/10.1029/2004GL020315
Aurela, M., Riutta, T., Laurila, T., Tuovinen, J.-P., Vesala, T., Tuittila, E.-S., et al. (2007). CO2 balance of a sedge fen in southern Finland—The
influence of a drought period. Tellus Series B, 59, 826–837.
Aurela, M., Tuovinen, J. P., & Laurila, T. (2001). Net CO2 exchange of a subarctic mountain birch ecosystem. Theoretical and Applied
Climatology, 70(1-4), 135–148. https://doi.org/10.1007/s007040170011
Barr, A. G., Richardson, A. D., Hollinger, D. Y., Papale, D., Arain, M. A., Black, T. A., et al. (2013). Use of change-point detection for friction-velocity
threshold evaluation in eddy-covariance studies. Agricultural and Forest Meteorology, 171, 31–45.
Brix, H., Sorrell, B. K., & Orr, B. T. (1992). Internal pressurization and convective gas-flow in some emergent fresh-water macrophytes.
Limnology and Oceanography, 37(7), 1420–1433. https://doi.org/10.4319/lo.1992.37.7.1420
Brown, M. G., Humphreys, E. R., Moore, T. R., Roulet, N. T., & Lafleur, P. M. (2014). Evidence for a nonmonotonic relationship between
ecosystem-scale peatland methane emissions and water table depth. Journal of Geophysical Research: Biogeosciences, 119, 826–835.
https://doi.org/10.1002/2013JG002576
Bubier, J., Moore, T., Savage, K., & Crill, P. (2005). A comparison of methane flux in a boreal landscape between a dry and a wet year.
Global Biogeochemical Cycles, 19, GB1023. https://doi.org/10.1029/2004GB002351
Chanton, J. P., Bauer, J. E., Glaser, P. A., Siegel, D. I., Kelley, C. A., Tyler, S. C., et al. (1995). Radiocarbon evidence for the substrates supporting
methane formation within northern Minnesota peatlands. Geochimica et Cosmochimica Acta, 59(17), 3663–3668. https://doi.org/10.1016/
0016-7037(95)00240-Z
Ciais, P., Sabine, C., Bala, G., Bopp, L., Brovkin, V., Canadell, J., et al. (2013). Carbon and other biogeochemical cycles. In T. F. Stocker, et al. (Eds.),
Climate change 2013: The physical science basis. Contribution of working group I to the fifth assessment report of the Intergovernmental Panel
on Climate Change (pp. 465–570). Cambridge, UK and New York: Cambridge University Press.
Couwenberg, J., & Fritz, C. (2012). Towards developing IPCC methane ‘emission factors’ for peatlands (organic soils). Mires and Peat, 10,
1–17.
Drebs, A., Nordlund, A., Karlsson, P., Helminen, J., & Rissanen, P. (2002). Climatological statistics of Finland 1971–2000, Finnish meteorological
institute, Helsinki, 99 ISBN 951–697–568-2.
Dunfield, P., Dumont, R., & Moore, T. R. (1993). Methane production and consumption in temperate and subarctic peat soils, response to
temperature and pH. Soil Biology and Biochemistry, 25(3), 321–326. https://doi.org/10.1016/0038-0717(93)90130-4
Foken, T., & Wichura, B. (1996). Tools for quality assessment of surface-based flux measurements. Agricultural and Forest Meteorology, 78(1-2),
83–105. https://doi.org/10.1016/0168-1923(95)02248-1
Fortuniak, K., Pawlak, W., Bednorz, L., Grygoruk, M., Siedlecki, M., & Zielinski, M. (2017). Methane and carbon dioxide fluxes of a temperate
mire in Central Europe. Agricultural and Forest Meteorology, 232, 306–318. https://doi.org/10.1016/j.agrformet.2016.08.023
Guenther, A. B., Jiang, X., Heald, C. L., Sakulyanontvittaya, T., Duhl, T., Emmons, L. K., & Wang, X. (2012). The model of emissions of gases and
aerosols from nature version 2.1 (MEGAN2.1): An extended and updated framework for modeling biogenic emissions. Geoscientific Model
Development, 5(6), 1471–1492. https://doi.org/10.5194/gmd-5-1471-2012
Hargreaves, K. J., Fowler, D., Pitcairn, C. E. R., & Aurela, M. (2001). Annual methane emission from Finnish mires estimated from eddy
covariance campaign measurements. Theoretical and Applied Climatology, 70(1-4), 203–213. https://doi.org/10.1007/s007040170015
Hari, P., & Kulmala, M. (2005). Station for measuring ecosystem-atmosphere relations (SMEAR II). Boreal Environment Research, 10, 315–322.
Hommeltenberg, J., Mauder, M., Drösler, M., Heidbach, K., Werle, P., & Schmid, H. P. (2014). Ecosystem scale methane fluxes in a natural
temperate bog-pine forest in southern Germany. Agricultural and Forest Meteorology, 198–199, 273–284.
Hui, D., Luo, Y., & Katul, G. (2003). Partitioning interannual variability in net ecosystem exchange between climatic variability and functional
change. Tree Physiology, 23(7), 433–442. https://doi.org/10.1093/treephys/23.7.433
IPCC (2013). Summary for Policymakers. In T. F. Stocker, et al. (Eds.), Climate change 2013: The physical science basis. Contribution of working
group I to the fifth assessment report of the Intergovernmental Panel on Climate Change (pp. 3–29). Cambridge, UK and New York:
Cambridge University Press.
Jackowicz-Korczyński, M., Christensen, T. R., Bäckstrand, K., Crill, P., Friborg, T., Mastepanov, M., & Ström, L. (2010). Annual cycle of methane
emission from a subarctic peatland. Journal of Geophysical Research, 115, G02009. https://doi.org/10.1029/2008JG000913
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1104
Acknowledgments
The study was supported by the
National Centre of Excellence (272041),
ICOS-Finland (281255), and CARB-ARC
(286190). T. Vesala was supported by an
Academy professor project (284701)
funded by the Academy of Finland, and
E.-S. Tuittila by Academy of Finland
project SA287039. We thank Pauli
Jokela at the Finnish Meteorological
Institute for climate data from Juupajoki
Hyytiälä station. We also thank the two
anonymous reviewers for their
thorough and constructive reviews.
The data are available at AVAA data
platform http://avaa.tdata.fi/web/
smart/smear.
Joabsson, A., & Christensen, T. R. (2001). Methane emissions fromwetlands and their relationship with vascular plants. Global Change Biology,
7(8), 919–932. https://doi.org/10.1046/j.1354-1013.2001.00044.x
Joabsson, A., Christensen, T. R., & Wallén, B. (1999). Vascular plant controls on methane emissions from northern peatforming wetlands.
TREE, 14(10), 385–388.
Juottonen, H., Tuittila, E.-S., Juutinen, S., Fritze, H., & Yrjälä, K. (2008). Seasonality of rDNA- and rRNA-derived archaeal communities and
methanogenic potential in a boreal mire. The ISME Journal, 2(11), 1157–1168. https://doi.org/10.1038/ismej.2008.66
Kim, J., Verma, S. B., & Billesbach, D. P. (1998). Seasonal variation in methane emission from a temperate Phragmites-dominated marsh: Effect
of growth stage and plant mediated transport. Global Change Biology, 5, 433–440.
Kim, J., Verma, S. B., Billesbach, D. P., & Clement, R. J. (1998). Diel variation in methane emission from a midlatitude prairie wetland:
Significance of convective throughflow in Phragmites australis. Journal of Geophysical Research, 103(D21), 28,029–28,039. https://doi.org/
10.1029/98JD02441
King, J. Y., & Reeburgh, W. S. (2002). A pulse-labeling experiment to determine the contribution of recent plant photosynthates to net
methane emission in arctic wet sedge tundra. Soil Biology and Biochemistry, 34(2), 173–180. https://doi.org/10.1016/
S0038-0717(01)00164-X
King, J. Y., Reeburgh, W. S., Thieler, K. K., Kling, G. W., Loya, W. M., Johnson, L. C., & Nadelhoffer, K. J. (2002). Pulse-labeling studies of carbon
cycling in Arctic tundra ecosystems: The contribution of photosynthates to methane emission. Global Biogeochemical Cycles, 16(4), 1062.
https://doi.org/10.1029/2001GB001456
Knox, S. H., Matthes, J. H., Sturtevant, C., Oikawa, P. Y., Verfaillie, J., & Baldocchi, D. (2016). Biophysical controls on interannual variability in
ecosystem-scale CO2 and CH4 exchange in a California rice paddy. Journal of Geophysical Research: Biogeosciences, 121, 978–1001. https://
doi.org/10.1002/2015JG003247
Kowalska, N., Chojnicki, B. H., Rinne, J., Haapanala, S., Siedlecki, P., Urbaniak, M., et al. (2013). Measurements of methane emission from a
temperate wetland by the eddy covariance method. International Agrophysics, 27, 283–290.
Laine, A., Bubier, J., Riutta, T., Nilsson, M., Moore, T., Vasander, H., & Tuittila, E.-S. (2012). Abundance and composition of plant biomass as
potential controls for mire NEE. Botany, 90(1), 63–74. https://doi.org/10.1139/b11-068
Lamers, L. P. M., Vile, M. A., Grootjans, A. P., Acreman, M. C., van Diggelen, R., Evans, M. G., et al. (2015). Ecological restoration of rich fens in
Europe and North America: from trial and error to an evidence-based approach. Biological Reviews, 90, 182–203. https://doi.org/10.1111/
brv.12102
Lees, K. J., Quaife, T., Artz, R. R. E., Khomik, M., & Clark, J. M. (2018). Potential for using remote sensing to estimate carbon fluxes across
northern peatlands—A review. Science of the Total Environment, 615, 857–874.
Levy, P. E., & Gray, A. (2015). Greenhouse gas balance of a semi-natural peatbog in northern Scotland. Environmental Research Letters, 10, 094019.
Li, T., Raivonen, M., Alekseychik, P., Aurela, M., Lohila, A., Zheng, X., et al. (2016). Importance of vegetation classes in modeling CH4 emissions
from boreal and subarctic wetlands in Finland. Science of the Total Environment, 572, 1111–1122. https://doi.org/10.1016/
j.scitotenv.2016.08.020
Lloyd, J., & Taylor, J. A. (1994). On the temperature-dependence of soil respiration. Functional Ecology, 8(3), 315–323. https://doi.org/10.2307/
2389824
Mammarella, I., Launiainen, S., Gronholm, T., Keronen, P., Pumpanen, J., Rannik, Ü., & Vesala, T. (2009). Relative humidity effect on the high
frequency attenuation of water vapour flux measured by a closed-path eddy covariance system. Journal of Atmospheric and Oceanic
Technology, 26(9), 1856–1866. https://doi.org/10.1175/2009JTECHA1179.1
Mammarella, I., Peltola, O., Nordbo, A., Järvi, L., & Rannik, Ü. (2016). Quantifying the uncertainty of eddy covariance fluxes due to the use of
different software packages and combinations of processing steps in two contrasting ecosystems. Atmospheric Measurement Techniques,
9(10), 4915–4933. https://doi.org/10.5194/amt-9-4915-2016
Marushchak, M. E., Friborg, T., Biasi, C., Herbst, M., Johansson, T., Kiepe, I., et al. (2016). Methane dynamics in the subarctic tundra: Combining
stable isotope analyses, plot- and ecosystem-scale flux measurements. Biogeosciences, 13(2), 597–608. https://doi.org/10.5194/
bg-13-597-2016
Mathijssen, P., Väliranta, M., Korrensalo, A., Alekseychik, P., Vesala, T., Rinne, J., & Tuittila, E.-S. (2016). Reconstruction of Holocene carbon
dynamics in a large boreal peatland complex, southern Finland. Quaternary Science Reviews, 142, 1–15.
Mathijssen, P. J. H., Kähkölä, N., Tuovinen, J.-P., Lohila, A., Minkkinen, K., Laurila, T., & Väliranta, M. (2017). Lateral expansion and carbon
exchange of a boreal peatland in Finland resulting in seven thousand years of positive radiative forcing. Journal of Geophysical Research:
Biogeosciences, 122, 562–577. https://doi.org/10.1002/2016JG003749
McVeigh, P., Sottocornola, M., Foley, N., Leahy, P., & Kiely, G. (2014). Meteorological and functional response partitioning to explain
interannual variability of CO2 exchange at an Irish Atlantic blanket bog. Agricultural and Forest Meteorology, 194, 8–19. https://doi.org/
10.1016/j.agrformet.2014.01.017
Mikhaylov, O. A., Miglovets, M. N., & Zagirova, S. V. (2015). Vertical methane fluxes in mesooligotrophic boreal peatland in European
Northeast Russia. Contemporary Problems of Ecology, 8(3), 368–375. https://doi.org/10.1134/S1995425515030099
Nadeau, D. F., Rousseau, A. N., Coursolle, C., Margolis, H. A., & Parlange, M. B. (2013). Summer methane fluxes from a boreal bog in northern
Quebec, Canada, using eddy covariance measurements. Atmospheric Environment, 81, 464–474. https://doi.org/10.1016/
j.atmosenv.2013.09.044
Oikawa, P. Y., Jenerette, G. D., Knox, S. H., Sturtevant, C., Verfaillie, J., Dronova, I., et al. (2017). Evaluation of a hierarchy of models reveals
importance of substrate limitation for predicting carbon dioxide and methane exchange in restored wetlands. Journal of Geophysical
Research: Biogeosciences, 122, 145–167. https://doi.org/10.1002/2016JG003438
Olson, D. M., Griffis, T. J., Noormets, A., Kolka, R., & Chen, J. (2013). Interannual, seasonal, and retrospective analysis of the methane and
carbon dioxide budgets of a temperate peatland. Journal of Geophysical Research: Biogeosciences, 118, 226–238. https://doi.org/10.1002/
jgrg.20031
Peltola, O., Hensen, A., Helfter, C., Belelli Marchesini, L., Bosveld, F. C., van den Bulk, W. C. M., et al. (2014). Evaluating the performance of
commonly used gas analysers for methane eddy covariance flux measurements: The InGOS inter-comparison field experiment.
Biogeosciences, 11(12), 3163–3186. https://doi.org/10.5194/bg-11-3163-2014
Peltola, O., Mammarella, I., Haapanala, S., Burba, G., & Vesala, T. (2013). Field intercomparison of four methane gas analyzers suitable for eddy
covariance flux measurements. Biogeosciences, 10(6), 3749–3765. https://doi.org/10.5194/bg-10-3749-2013
Petrescu, A. M. R., Lohila, A., Tuovinen, J.-P., Baldocchi, D., Desai, A. R., Roulet, N. T., et al. (2015). The uncertain climate footprint of wetlands
under human pressure. Proceedings of the National Academy of Sciences, 112(15), 4594–4599. https://doi.org/10.1073/pnas.1416267112
Raivonen, M., Smolander, S., Backman, L., Susiluoto, J., Aalto, T., Markkanen, T., et al. (2017). HIMMELI v1.0: HelsinkI Model of MEthane
buiLd-up and emIssion for peatlands. Geoscientific Model Development, 10(12), 4665–4691. https://doi.org/10.5194/gmd-10-4665-2017
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1105
Rinne, J., Aurela, M., & Manninen, T. (2009). A simple method to determine the timing of snowmelt by remote sensing with application to the
CO2 balances of northern mire and heath ecosystems. Remote Sensing, 1(4), 1097–1107. https://doi.org/10.3390/rs1041097
Rinne, J., Riutta, T., Pihlatie, M., Aurela, M., Haapanala, S., Tuovinen, J.-P., et al. (2007). Annual cycle of methane emission from a boreal fen
measured by the eddy covariance technique. Tellus Series B, 59(3), 449–457. https://doi.org/10.1111/j.1600-0889.2007.00261.x
Riutta, T., Laine, J., Aurela, M., Rinne, J., Vesala, T., Laurila, T., et al. (2007). Spatial variation in plant communities and their function regulates
carbon gas dynamics in boreal fen ecosystem. Tellus Series B, 59(5), 838–852. https://doi.org/10.1111/j.1600-0889.2007.00302.x
Roulet, N. T., Lafleur, P. M., Richard, P. J. H., Moore, T. R., Humpreys, E. R., & Bubier, J. (2007). Contemporary carbon balance and late Holocene
carbon accumulation in a northern peatland. Global Change Biology, 13(2), 397–411. https://doi.org/10.1111/j.1365-2486.2006.01292.x
Saarinen, T. (1996). Biomass and production of two vascular plants in a boreal mesotrophic fen. Canadian Journal of Botany, 74(6), 934–938.
https://doi.org/10.1139/b96-116
Ström, L., Ekberg, A., Mastepanov, M., & Christensen, T. R. (2003). The effect of vascular plants on carbon turnover and methane emissions
from a tundra wetland. Global Change Biology, 9(8), 1185–1192. https://doi.org/10.1046/j.1365-2486.2003.00655.x
Ström, L., Falk, J. M., Skov, K., Jackowicz-Korczynski, M., Mastepanov, M., Christensen, T. R., et al. (2015). Controls of spatial and temporal
variability in CH4 flux in a high arctic fen over three years. Biogeochemistry, 125(1), 21–35. https://doi.org/10.1007/s10533-015-0109-0
Suni, T., Rinne, J., Reissell, A., Altimir, N., Keronen, P., Rannik, Ü., et al. (2003). Long-term measurements of surface fluxes above a Scots pine
forest in Hyytiälä, southern Finland, 1996–2001. Boreal Environment Research, 8, 287–301.
Teklemariam, T. A., LaFleur, P. M., Moore, T. R., Roulet, N. T., & Humphreys, E. R. (2010). The direct and indirect effects of inter-annual
meteorological variability on ecosystem carbon dioxide exchange at a temperate ombrotrophic bog. Agricultural and Forest Meteorology,
150(11), 1402–1411. https://doi.org/10.1016/j.agrformet.2010.07.002
Tian, H., Xu, X., Liu, M., Ren, W., Zhang, C., Chen, G., & Lu, C. (2010). Spatial and temporal patterns of CH4 and N2O fluxes in terrestrial
ecosystems of North America during 1979–2008: Application of a global biogeochemistry model. Biogeosciences, 7(9), 2673–2694. https://
doi.org/10.5194/bg-7-2673-2010
Turetsky, M. R., Kotowska, A., Bubier, J., Dise, N. B., Crill, P., Hornibrook, E., et al. (2014). A synthesis of methane emissions from 71 northern,
temperate, and subtropical wetlands. Global Change Biology, 20(7), 2183–2197. https://doi.org/10.1111/gcb.12580
Tyndall, J. (1862). On the absorption and radiation of heat by gaseous matter—Second memoir. Philosophical Transactions. Royal Society of
London, 1862(152), 59–98.
van den Berg, M., Ingwersen, J., Lamers, M., & Streck, T. (2016). The role of Phragmites in the CH4 and CO2 fluxes in a minerotrophic peatland
in Southwest Germany. Biogeosciences, 13(21), 6107–6119. https://doi.org/10.5194/bg-13-6107-2016
Volta, D. A. (1777). Lettere del Signor don Alessandro Volta Sull Aria Inflammabile Nativa delle Paludi [Letters of Signor don Alessandro Volta
on the flammable native air of the marshes], Giuseppe Marelli, Milan, Italy.
Walter, B., & Heimann, M. (2000). A process-based climate-sensitive model to derive methane emissions from natural wetlands: Application
to five wetland sites, sensitivity to model parameters, and climate. Global Biogeochemical Cycles, 14(3), 745–765. https://doi.org/10.1029/
1999GB001204
Wang, H., Richardson, C. J., & Ho, M. (2015). Dual controls on carbon loss during drought in peatlands. Nature Climate Change, 5(6), 584–587.
https://doi.org/10.1038/nclimate2643
Wania, R., Ross, I., & Prentice, I. C. (2010). Implementation and evaluation of a new methane model within a dynamic global vegetation
model: LPJ-WHyMe v.1.3.1. Geoscientific Model Development, 3(2), 565–584. https://doi.org/10.5194/gmd-3-565-2010
Watts, J. D., Kimball, J. S., Parmentier, F.-J. W., Sachs, T., Rinne, J., Zona, D., et al. (2013). A satellite data driven biophysical modeling approach
for estimating northern peatland and tundra CO2 and CH4 fluxes. Biogeosciences, 11(7), 1961–1980. https://doi.org/10.5194/
bg-11-1961-2014
Webb, E. K., Pearman, G. I., & Leuning, R. (1980). Correction of flux measurements for density effects due to heat and water vapour transport.
Quarterly Journal of the Royal Meteorological Society, 106(447), 85–100. https://doi.org/10.1002/qj.49710644707
Whiting, G. J., & Chanton, J. P. (1992). Plant-dependent CH4 emission in a subarctic Canadian fen. Global Biogeochemical Cycles, 6(3), 225–231.
https://doi.org/10.1029/92GB00710
Whiting, G. J., & Chanton, J. P. (1993). Primary production control of methane emission from wetlands. Nature, 364(6440), 794–795. https://
doi.org/10.1038/364794a0
Whiting, G. J., Chanton, J. P., Bartlet, D. S., & Happell, J. D. (1991). Relationships between CH4 emission, biomass, and CO2 exchange in a
subtropical grassland. Journal of Geophysical Research, 96(D7), 13067–13,071. https://doi.org/10.1029/91JD01248
Wille, C., Kutzbach, L., Sachs, T., Wagner, D., & Pfeiffer, E.-M. (2008). Methane emission from Siberian arctic polygonal tundra: Eddy covariance
measurements and modeling. Global Change Biology, 14(6), 1395–1408. https://doi.org/10.1111/j.1365-2486.2008.01586.x
Wilson, D., Alm, J., Riutta, T., Laine, J., Byrne, K. A., Farrell, E. P., & Tuittila, E.-S. (2007). A high resolution green area index for modelling the
seasonal dynamics of CO2 exchange in vascular plant peatland communities. Plant Ecology, 190(1), 37–51. https://doi.org/10.1007/
s11258-006-9189-1
Zhuang, Q., Melillo, J. M., Kicklighter, D. W., Prinn, R. G., McGuire, A. D., Steudler, P. A., et al. (2004). Methane fluxes between terrestrial
ecosystems and the atmosphere at northern high latitudes during the past century: A retrospective analysis with a process-based
biogeochemistry model. Global Biogeochemical Cycles, 18, GB3010. https://doi.org/10.1029/2004GB002239
10.1029/2017GB005747Global Biogeochemical Cycles
RINNE ET AL. 1106