A decade of boreal rich fen greenhouse gas fluxes inresponse to natural and experimental water tablevariabilityDAV ID OLEFELDT 1 , 2 , EUG �EN IE S . EUSK IRCHEN3 , J ENN I FER HARDEN4 , EVAN KANE 5 ,

A . DAV ID MCGU IRE 6 , MARK P . WALDROP 4 and MERRITT R. TURETSKY1

1Department of Integrative Biology, University of Guelph, Science Complex, Guelph, ON N1G 2W1, Canada, 2Department of

Renewable Resources, University of Alberta, Edmonton, AB T6G 2H1, Canada, 3Institute of Arctic Biology, University of Alaska

Fairbanks, Fairbanks, AK 99775, USA, 4U.S. Geological Survey, Menlo Park, CA 94025, USA, 5School of Forest Resources and

Environmental Sciences, and USDA Forest Service, Michigan Tech University, Northern Research Station, Houghton, MI 49931,

USA, 6U.S. Geological Survey, Alaska Cooperative Fish and Wildlife Research Unit, University of Alaska Fairbanks, Fairbanks,

AK 99775, USA

Abstract

Rich fens are common boreal ecosystems with distinct hydrology, biogeochemistry and ecology that influence their

carbon (C) balance. We present growing season soil chamber methane emission (FCH4), ecosystem respiration (ER),

net ecosystem exchange (NEE) and gross primary production (GPP) fluxes from a 9-years water table manipulation

experiment in an Alaskan rich fen. The study included major flood and drought years, where wetting and drying

treatments further modified the severity of droughts. Results support previous findings from peatlands that drought

causes reduced magnitude of growing season FCH4, GPP and NEE, thus reducing or reversing their C sink function.

Experimentally exacerbated droughts further reduced the capacity for the fen to act as a C sink by causing shifts in

vegetation and thus reducing magnitude of maximum growing season GPP in subsequent flood years by ~15% com-

pared to control plots. Conversely, water table position had only a weak influence on ER, but dominant contribution

to ER switched from autotrophic respiration in wet years to heterotrophic in dry years. Droughts did not cause inter-

annual lag effects on ER in this rich fen, as has been observed in several nutrient-poor peatlands. While ER was

dependent on soil temperatures at 2 cm depth, FCH4 was linked to soil temperatures at 25 cm. Inter-annual variability

of deep soil temperatures was in turn dependent on wetness rather than air temperature, and higher FCH4 in flooded

years was thus equally due to increased methane production at depth and decreased methane oxidation near the sur-

face. Short-term fluctuations in wetness caused significant lag effects on FCH4, but droughts caused no inter-annual

lag effects on FCH4. Our results show that frequency and severity of droughts and floods can have characteristic

effects on the exchange of greenhouse gases, and emphasize the need to project future hydrological regimes in rich

fens.

Keywords: carbon dioxide, climate change, ecosystem respiration, methane, peatland, soil temperature, water table, wetland

Received 7 April 2016; revised version received 6 November 2016 and accepted 7 December 2016

Introduction

Northern peatlands cover ~3% of the global land cover

and are dominant ecosystems in many boreal regions.

As the end of the last glaciation, peatlands have accumu-

lated ~500 Tg carbon (C) in the form of peat (Yu, 2012).

This represents ~15–30% of the total current global soil C

pool (Batjes, 1996). Accumulation of soil C in northern

peatlands is primarily a result of restricted rates of

decomposition under cool and often anaerobic soil con-

ditions (Clymo et al., 1998; Roulet et al., 2007; Yu, 2012).

However, anaerobic conditions also promote the

production and release of methane (CH4), and northern

peatlands are responsible for ~20% of all natural CH4

sources to the atmosphere (Bergamschi et al., 2007).

Although CH4 is a more potent greenhouse gas than

CO2, it has a much shorter half-life in the atmosphere

(Hartmann et al., 2013). The net effect of sustained CO2

uptake and CH4 release from northern peatlands over

the Holocene has overall resulted in a net cooling effect

on the global climate system (Frolking & Roulet, 2007).

The future greenhouse gas exchange of northern peat-

lands is uncertain, but will likely be strongly influenced

by interactions between peatland type and climate

change impacts on hydrological regimes.

Recent climate change at high latitudes has been

occurring at rates faster than the global average. InteriorCorrespondence: David Olefeldt, tel. 780 248 1814, fax 780 492 4323,

e-mail: olefeldt@ualberta.ca

2428 © 2017 John Wiley & Sons Ltd

Global Change Biology (2017) 23, 2428–2440, doi: 10.1111/gcb.13612

Alaska has over the last few decades experienced both

increasing air temperatures and an amplification of the

hydrological cycle that includes increases in precipita-

tion, potential evapotranspiration and river discharge

(Hinzman et al., 2005; Serreze & Francis, 2006; Wendler

& Shulski, 2009; Rawlins et al., 2010). An amplified

hydrological cycle is likely to redistribute soil moisture

at the landscape scale and is expected to cause reduced

summer soil moisture conditions in ecosystems that are

largely dependent on precipitation inputs (Rouse, 1998;

Lafleur et al., 2005; Berg et al., 2009). Rich fens, which

receive substantial water inputs from their surrounding

landscapes, could, however, experience a differential

response due to potentially coinciding greater surface

water and groundwater runoff (Walvoord & Striegl,

2007; Olefeldt & Roulet, 2012; Tardif et al. 2015).

The position of the water table in a peatland often

has strong influences on the greenhouse gas exchange.

A higher water table is generally associated with higher

net methane emissions (FCH4), as the balance between

anaerobic methane production and aerobic oxidation is

shifted, but vegetation composition can modify this

relationship significantly (Segers, 1998; Limpens et al.,

2008; Olefeldt et al., 2013; Turetsky et al., 2014). Differ-

ent peatland ecosystems have also shown positive, neg-

ative and no relationships between ecosystem

respiration (ER) and water table position (Chimner &

Cooper, 2003; Lafleur et al., 2005; Ballantyne et al., 2014;

Juszczak et al., 2013; McConnell et al., 2013). This is

potentially due to independent responses of constituent

autotrophic and heterotrophic respiration. Peat miner-

alization, i.e., heterotrophic respiration, increases sub-

stantially under aerobic conditions (Moore & Knowles,

1989; Silvola et al., 1996). However, droughts can also

influence gross primary productivity (GPP) of wetland

plant species (Sulman et al., 2009; Adkinson et al., 2011;

Lund et al., 2012) and thus affect rates of autotrophic

respiration (Crow & Wieder, 2005; Han et al., 2014).

Vegetation composition further influences the location

of optimal water table position for maximum GPP and

net ecosystem exchange (NEE) (Yurova et al., 2007;

Adkinson et al., 2011). Altered frequencies or severities

of droughts and floods are thus likely to affect the C

sink function of peatlands, but responses are also likely

to depend on peatland type.

Droughts and floods can further impact peatland C

balance through lag effects over various timescales.

Water table fluctuations can cause short-term lag effects

on FCH4 through transient soil conditions, including the

regeneration or depletion of terminal electron acceptors

and re-establishment of microbial communities (Dow-

rick et al., 2006; Knorr & Blodau, 2009; Deppe et al.,

2010; Sun et al., 2012). In nutrient-poor peatlands, the

degradation of phenolic compounds during severe

droughts has been shown to enable drastically

increased rates of anaerobic heterotrophic respiration

in subsequent wet years (Fenner & Freeman, 2011).

Over longer timescales, it is likely that the most impor-

tant effects of altered drought and flood characteristics

are due to induced shifts in vegetation composition, as

individual vegetation communities have specific rela-

tionships between C fluxes and abiotic variables such

as water table position, soil temperatures and light con-

ditions (Laiho, 2006; Lindroth et al., 2007; Olefeldt et al.,

2013; Ward et al., 2013).

The Alaska Peatland Experiment (APEX) was initi-

ated in a rich fen in 2005 as a long-term ecosystem-scale

experiment designed to study potential effects of cli-

mate change on peatland greenhouse gas exchange.

One goal at APEX was to create a lowered and a raised

water table regime through water table manipulations,

yet without altering the natural inter- and intra-annual

water table variability that characterizes these ecosys-

tems. In this study, our objective is to investigate the

influences of water table position and variability on rich

fen C fluxes, including long-term influences arising due

to experimentally altered drought severity.

Materials and methods

Study site and experimental design

The Alaska Peatland Experiment is located adjacent to the

Bonanza Creek Long-Term Ecological Research forest, ~35 km

southwest of Fairbanks, Alaska, USA (64.82°N, 147.87 W).

Mean annual temperature (1917–2000) is �3.1 °C, and mean

annual precipitation is 287 mm (Hinzman et al., 2006). The site

is positioned within the Tanana River floodplain and is char-

acterized as a rich fen (surface water pH 5.2–5.4), with vegeta-

tion dominated by marsh cinquefoil (Potentilla palustris),

wheat sedge (Carex atherodes), water horsetail (Equisetum fluvi-

atile) and a ground cover mostly comprised of brown mosses

(Drepanocladus aduncus and Hamatocaulis vernicosus) and sparse

Sphagnum spp. (Churchill et al., 2015). Biomass harvest indi-

cates an aboveground net primary productivity of vascular

plants of ~300 g m�2 yr�1. A maximum vascular green area of

~2.5 m2 m�2 is attained between late June and mid-August

(Churchill et al., 2015). Peat depth is ~1 m, the site lacks per-

mafrost, and it has no distinct microtopography.

In the spring of 2005, three 120-m2 plots were randomly

assigned to a control, raised and lowered water table treat-

ment. Drainage channels were dug around the lowered water

table plot to divert water ~20 m downslope to a surface well,

from which solar-powered bilge pumps added up to

100 mm day�1 to the raised plot during the thawed seasons

(for further information, see Turetsky et al., 2008). No signifi-

cant difference in water table position or vegetation composi-

tion among plots was observed prior to treatment initiation.

Within each plot, six subplots were established for greenhouse

gas flux measurements.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2429

Water table and temperature records

Data loggers (CR10x, Campbell Scientific, Logan, UT)

recorded hourly air temperature at 1.5 m height, hourly water

table positions in 1-m-long 5-cm-diameter PVC wells in each

treatment plot and hourly soil temperatures at 2 and 25 cm

depths in all 18 subplots and at 50 cm depth in six subplots.

Daily data records were compiled using the hourly data.

There were several gaps in the continuous temperature and

water table records due to sensor or logger malfunctions dur-

ing the 9-year study period. Soil temperature records had 75%

data coverage on average, ranging between 23 and 90% cover-

age for individual sensors. Variation in soil temperature

among subplots was not associated with water table treat-

ments, or an intended warming treatment (see supplementary

information; Fig. S1). Hence, we compiled single site common

soil temperature records for each depth by using the average

of daily soil temperatures across subplots (Fig. 1a). Weekly

manual measurements of water table position within each plot

were used to linearly gap-fill the continuous records during

parts of 2006 (lowered and control plots) and for all of 2007

(all plots) (Fig. 1b).

Measuring greenhouse gas fluxes

Greenhouse gas flux measurements were taken using static

chamber techniques (Carrol & Crill, 1997). Collars (0.36 m2)

were inserted to 10 cm depth at all 18 subplots in 2005. A clear

chamber (0.23 m3) was constructed out of 0.6-cm-thick Lexan,

and an airtight seal was created between base and chamber

using foam tape applied during each measurement campaign.

Two internal fans were used to mix the air within the chamber

during measurements.

Fluxes of CO2 were measured under ambient light condi-

tions (measuring NEE) followed by dark conditions using a

dark shroud (ER measurement). The difference between NEE

and ER equals our GPP estimate. Chambers were closed for 2–3 min, and CO2 concentrations were determined every 1.6 s

using a PP Systems EGM-4 portable infrared gas analyzer

(IRGA; Amesbury, MA, USA). The IRGA was calibrated

before each measurement campaign, using external CO2 stan-

dards. From 2006 and onward, a PP Systems TRP-1 measured

temperature and photosynthetic photon flux density (PPFD,

lmol m�2 s�1) within the chamber. Chamber measurements

of CH4 were typically taken on days immediately following

CO2 measurements due to time constraints. Chambers were

closed for 30–40 min, and four 20 mL gas samples were taken,

using plastic syringes with three-way stop cocks. Samples

were analyzed within 24 h, using a Varian 3800 gas chro-

matograph with a FID detector with a Haysep N column (Var-

ian Analytical Inc., Palo Alto, California). We report net CO2

(reported in lmol CO2 m�2 s�1) and net CH4 (reported in mg

CH4 m�2 day�1) fluxes to the atmosphere as positive and net

uptake as negative.

Sampling was initiated between May 20 and June 15 each

year except during 2012 and 2013 when the site was flooded

and measurements could not start until July 1 and August 5,

respectively. Weekly sampling was carried out in 2005–2007and 2010–2013, while 2008 and 2009 had biweekly sampling.

Last samplings were carried out between September 1 and

October 1 except in 2008 and 2009 when they were completed

in mid-July. A total of 1380 paired NEE and ER flux measure-

ments and 918 FCH4 measurements were accepted after data

quality check (see supplementary information). Maximum

and minimum measurements per year were 259 and 46 for

2005 and 2013, respectively, for CO2 and 219 and 31 in 2006

and 2013, respectively, for FCH4.

We define fluxes measured between day-of-year (DOY) 165

and 235 (June 13 to August 21 in non-leap years) as peak

growing season fluxes (NEEPeak, GPPPeak, ERPeak, FCH4Peak).

–75

–50

–25

0

25

50

Wat

er ta

ble

posi

tion

(cm

)

ControlLoweredRaised

0

5

10

15

20

25

Tem

pera

ture

(°C

) Air Temp. 2 cm 25 cm 50 cm

2005 2006 2007 2008 2009 2010 2011 2012 2013

(a)

(b)

Fig. 1 Peak growing season (June 13–August 22) data from 2005 to 2013, including (a) site common air and soil temperatures at 2, 25

and 50 cm and (b) water table position for each treatment plot (positive values indicate water table above peat surface).

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2430 D. OLEFELDT et al.

Peak growing season was defined to include the period of

maximum vascular green area (Churchill et al., 2015). As sam-

pling periods varied substantially among years, using peak

growing season data allows for a better comparison of abiotic

controls on C fluxes among years and treatments as it reduces

confounding influences that arise due to seasonally develop-

ing phenology (Peichl et al. 2015). Peak growing seasons

included 65% of CO2 flux measurements and 74% of CH4 mea-

surements (Figs S2 and S3). Whether all data or only peak

growing season data are used is explicitly stated for each anal-

ysis.

Data modeling and statistical analysis

We used both linear and nonlinear analyses to assess abiotic

controls on FCH4, ER, NEE and GPP. All statistical analyses

were conducted in MatLab R2014a, with the Statistics Toolbox

and the Curve Fitting Toolbox (v 3.4.1) (MathWorks, Natick,

Massachusetts). The linear mixed effects models (function:

fitlme) included a categorical variable for water table

treatment (control, raised, lowered) and continuous abiotic

variables as fixed effects. Collar ID was included as a ran-

dom effect nested within treatment plots to account for the

lack of independence of repeated measurements. Abiotic

variables included water table position (WT), soil tempera-

tures at 2 and 25 cm depth (T2 and T25) and PPFD (only

included for NEE and GPP analysis). Interactions between

water table treatments and all abiotic variables were also

included as fixed effects. The analysis was carried out on

log10-transformed FCH4 fluxes [log10(FCH4)], due to a non-

normal distribution. Logarithmic transformation excluded

negative FCH4 fluxes, representing 8% of FCH4 measure-

ments. Analysis of variance (ANOVA) was performed on the

marginal effects, and yielded F and P values for each fixed

effect, including interactions. Significant variables, as indi-

cated by the linear mixed effects model, were subsequently

included in a forward stepwise multiple linear regression

(function: stepwise) to estimate parameter coefficients for

each variable.

We used residuals from the stepwise linear model to assess

potential time lags in relationships between water table posi-

tion and ER and FCH4. Coefficients of determination were

determined for linear correlations between model residuals

and the net shift in water table position over a time period

preceding a flux measurement. Time periods for lag effects

ranging from 1 to 50 days were considered. Inter-annual time-

lag effects, i.e., the effect of the wetness of the preceding year

on the current year fluxes, were assessed by linear correlations

between the current year average model residual within each

plot and its average water table position during the preceding

year.

A nonlinear model was used to assess temperature

sensitivity:

Flux ¼ A�QðT10Þ10 ð1Þ

where Flux is either FCH4 or ER, A is Flux at 0 °C, Q10 is the

temperature dependence of Flux and T is soil temperature at

either 2 or 25 cm below the surface. The analysis was carried

out for parsed datasets to assess differences in temperature

dependence under four water table ranges; <�25 cm, �25 to

�10 cm, �10 to 0 cm and >0 cm (positive values indicate

water table above the surface).

Nonlinear dependence of GPP on variation in PPFD was

modeled using:

GPP ¼ GPPmax � PPFD

kþ PPFDð2Þ

where GPPmax is the maximum rate of GPP under light satura-

tion and k is the PPFD level where half of GPPmax is attained.

Our analysis estimated parameters using only peak growing

season GPP data, for six groups based on water table treat-

ment and annual wetness (dry years vs. wet years; see below).

Optimal water table position and the range over which

maximum fluxes occur were estimated as:

Flux ¼ Fbase þ Fopt � exp�0:5� ðWT�WToptÞ2

WT2rng

� �ð3Þ

where Flux is either log10(FCH4Peak), ERPeak, GPPPeak or

NEEPeak, Fbase is Flux outside the range of optimal water table

position, Fopt is the addition to Fbase at the optimal water table

position (i.e., Fbase + Fopt = maximum Flux), WT is the water

table position at the time of flux measurement, WTopt is the

water table position where maximal Flux occurs and WTrng is

the distance of the water table range around WTopt where

increased Flux occurs. Unit of Fbase and Fopt is the same as for

Flux, while the unit for WT, WTopt and WTrng is in cm. Equa-

tion 2 is a modified version of an equation used by Tuitilla

et al. (2004) and Chivers et al. (2009), but it assumes that fluxes

can be 6¼ 0 outside the water table range for optimal fluxes.

For NEE and GPP, we used only data measured when

PPFD > 400 lmol m�2 s�1, i.e., when light limitation was

minimal (see Results). Data from all water treatment plots

were pooled for this analysis, as data from each treatment sep-

arately did not yield significant model parameters.

Results

Climate and abiotic variables

During our measurement period, mean annual air tem-

peratures ranged from �1.7 °C (2005) to �4.9 °C (2012),

while peak growing season average air temperatures

varied between 14.4 °C (2008) to 17.0 °C (2013)

(Fig. 1a). Average peak growing season soil tempera-

tures were more variable than air temperatures, and

inter-annual variability increased with soil depth

(Fig. 1a). The standard deviation of average growing

season temperatures over the nine study years was

0.9 °C for air temperature, 1.1 °C for soil temperature

at 2 cm, 2.0 °C at 25 cm and 2.8 °C at 50 cm. Average

peak growing season soil temperature at 2 cm was sig-

nificantly correlated with average peak growing season

air temperatures, while soil temperature at 25 cm was

significantly correlated with average peak growing

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2431

season water table position (Fig. S4). The average peak

growing season temperature at 25 cm depth was ~8and ~12 °C in dry and wet years (see below), respec-

tively (Fig. 1a, b).

Water table position varied greatly between years,

with average peak growing season water table posi-

tions in the control plot varying between �32 cm (2006)

and +21 cm (2013) (Fig. 1b). In subsequent analysis, we

define years when the control plot had average grow-

ing season water table position below �20 cm as dry

years (2006, 2010 and 2011) and other years as wet

years. The water table treatments had no effect on

water table position among treatments in wet years, but

in dry years the lowered plot had an average water

table position that was 9 cm lower than the control and

the raised plot had an average water table position

11 cm higher than the control plot (Fig. 1b).

Methane fluxes

The linear model found measured log-transformed

FCH4 fluxes [log10(FCH4)] to be strongly related to WT

and T25, with minor, but significant, influences from

both T2 and the interaction between water table treat-

ment and WT (Table S1; overall R2adj = 0.63). The non-

linear model showed that variability of T25 explained

between 28 and 33% of the variability in FCH4, except

for during the driest periods (WT < �25 cm) when

only 10% of variability was explained (Fig. 2a,

Table S2). Variability in T2 only explained between 4%

and 9% of FCH4 (Table S2). The seasonal trend in FCH4

during wet years thus followed the seasonal trend of

T25, leading to a late peak emissions period between

mid-August and late September (Fig. S2). Higher water

table led to higher FCH4 (Fig. 2a), and maximum

log10(FCH4) was modeled (Eqn 3) to occur when the

water table was well above the peat surface

(WTopt = 8.0 � 3.4 cm) (Table 1). As a result of the

influences of wetness and T25, wet years with associ-

ated warmer T25 (see above) had 4–20 times greater

average FCH4Peak than dry years associated with colder

T25 (Fig. 3a).

Significant short-term lags were present in the rela-

tionship between FCH4 and water table position. The

residuals from the linear model were significantly

correlated with the net shift in water table position

that occurred 3–7 days prior to FCH4 measurements

(maximum R2 = 0.08, P < 0.01, Fig. 4). For example,

this lag effect indicated that measured FCH4 was

~25% higher than predicted by the linear model

when the water table had been dropping by 5 cm

over the last 5 days, and equally lower than pre-

dicted when the water table had been rising. We

found no evidence of inter-annual lag effects, as

residuals in average annual FCH4 were not signifi-

cantly related to the average water table position of

the preceding year in any plot.

Ecosystem respiration

The linear mixed effects model indicated significant

influences on measured ER fluxes from T2, T25, WT and

the interaction between T2 and treatment, but the

model explained only 20% of the variation in ER

(Table S1). The most important predictor of higher ER

was increasing T2 (Fig. 2b). The general dependency of

0

50

100

150

200

250

Met

hane

flux

(mg

CH

4 d–

1 )

Water table interval (cm)

3-66-99-1212-15

0

1

2

3

4

5

Eco

syst

em re

spira

tion

(μm

ol C

O2

m–2

s–1

)

Water table interval (cm)

4-1010-1515-21

Soil T2 (˚C)

Soil T25 (˚C)

(a) (b)

Fig. 2 Influence of water table position and soil temperature on a) measured methane fluxes and b) measured ecosystem respiration.

Data from both outside and within peak growing season are included. Bars represent median fluxes measured within specified water

table and soil temperature intervals. Positive water table position indicates water table above the peat surface. Error bars represent first

and third quartiles of the data. Note that soil temperatures at 25 and at 2 cm were used for methane fluxes and ecosystem respiration,

respectively, as data at these depths had the greatest explanatory power (see Table S2).

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2432 D. OLEFELDT et al.

ER on T2 led to a seasonal pattern with a period of max-

imum fluxes between June ~20 and August 10 (Fig. S2),

i.e., a longer and earlier peak period than for maximum

FCH4 (see above). Nonlinear analysis showed that T2

was a better predictor than T25 for ER, except under the

driest conditions when the water table was below

�25 cm (Table S2). The interactive effect of the water

table treatment and T2 on ER suggested that the low-

ered water table treatment plot had lower temperature

sensitivity of ER than the other plots – although this

effect may be an indirect consequence of treatment

effect on NEE/GPP through autotrophic respiration

Table 1 Estimated parameters for models examining the nonlinear effects of water table position for measured FCH4 and CO2

fluxes during peak growing season (Eqn 3). Fbase indicates the flux rates outside the water table optima, Fadd the increase in flux at

the water table optima, WTopt the position of the water table optima and WTrng the width of the water table optima. For WTopt, pos-

itive values represent water table positions above the peat surface. Parameters with 95% confidence bounds are presented. All mod-

els were significant (P < 0.01). Models applied to measured NEE and estimated GPP included only measurements taken under high

light conditions (PPFD > 400 lmol m�2 s�1)

Fbase Fadd WTrng WTopt R2adj

Methane flux

(mg CH4 m�2 day�1)

(mg CH4 m�2 day�1) (mg CH4 m�2 day�1) (cm) (cm)

Log10(FCH4Peak) 0.76 � 0.11 1.13 � 0.13 17.4 � 3.9 8.0 � 3.8 0.43

CO2 fluxes

(lmol CO2 m�2 s�1)

(lmol CO2 m�2 s�1) (lmol CO2 m�2 s�1) (cm) (cm)

ERPeak 2.97 � 0.07 1.31 � 0.22 5.4 � 2.1 �5.2 � 0.4 0.05

NEEPeak 0.30 � 0.81 �3.54 � 0.77 21.9 � 6.6 1.3 � 3.8 0.29

GPPPeak �3.34 � 0.86 �3.23 � 0.86 17.1 � 6.6 �1.8 � 4.0 0.14

–8

–6

–4

–2

0Wet Years Dry Years

Gro

ss p

rimar

y pr

oduc

tion

(μm

olC

O2

m–2

s–1

)

–4

–3

–2

–1

0Wet Years Dry Years

Net

eco

syst

em e

xcha

nge

(μm

olC

O2

m–2

s–1

)

0

1

2

3

4Wet Years Dry Years

Eco

syst

em re

spira

tion

(μm

ol C

O2

m–2

s–1

)

Control

LoweredRaised

a ab ab c d b

a a a a a a

a b ab bc c ab

0

25

50

75

100

125Wet years Dry Years

Met

hane

flux

(mg

CH

4 d–

1 )

a a a

c c b

(a) (b)

(c) (d)

Fig. 3 Average peak growing season (13 June–22 August) fluxes measured under high light conditions (PPFD > 400 lmol m�2 s�1) of

(a) methane fluxes, (b) ecosystem respiration, (c) net ecosystem exchange and (d) estimated gross primary production. Error bars repre-

sent �2 standard errors. Lowercase letters indicate significant differences between treatments based on a two-way ANOVA, followed by

a multiple comparison using a Bonferroni correction. Dry years were 2006, 2010 and 2011, and wet years were 2005, 2007, 2008, 2009,

2012 and 2013.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2433

rates (see below) which were not accounted for in this

linear model (Table S1).

Links between ER and WT were complex and not

easily captured by either the linear or nonlinear

approaches (Fig. 2b). The linear model found ER to

decrease marginally with a higher WT position

(Table S1), while the nonlinear model indicated that ER

was on average ~30% higher when the water table was

just below the surface (WTopt = �5.4 � 0.9 cm) com-

pared to positions higher or lower (Table 1). While this

nonlinear relationship was consistent across all experi-

mental treatments, it had low explanatory power

(R2adj = 0.04). Qualitatively, ER appeared to have two

optima, with maximum ER when the water table was

~10 cm and >40 cm below the surface (Fig. 2b). Overall,

we found that average ERPeak was not significantly

affected by the water table treatment and that it did not

vary between wet and dry years (Fig. 3b).

Water table fluctuations were found to cause neither

short-term (1–40 days) nor inter-annual lag effects on

ER. The residuals from our linear ER model (Table S1),

using data from all periods, were not significantly

related to net shifts in water table position over 1–40 days preceding the flux measurements (Fig. 4). The

average annual model residual was neither related to

the average water table position of the preceding year

in any plot.

During wet years, we found that average ERPeak of

individual subplots had a significant relationship with

NEEPeak measured under high light conditions

(>400 lmol m�2 s�1), suggesting that higher plant

productivity among subplots was associated with

higher ERPeak (Fig. 5a). The slope of this relationship

shows that variability among subplot NEEPeak was

associated with shifts in ERPeak representing

53 � 26% (95% CI) of the magnitude of shifts in

NEEPeak. This implies that among subplots, a

1.0 lmol m�2 s�1 higher average GPPPeak is asso-

ciated with an increase in average ERPeak by

0.35 � 0.14 (95% CI) lmol m�2 s�1. During dry years,

we found that subplots of the raised water table treat-

ment maintained the relationship between average

ERPeak and NEEPeak observed in wet years, but that

average ERPeak of control and lowered water table

treatment subplots became unrelated to NEEPeak

(Fig. 5a). Consequently, the ratio –NEEPeak/ERPeak

(relative magnitude of NEEPeak to ERPeak) was found

to be influenced by water table position among years,

with the ratio starting to drop once the average peak

growing season water table position was below

�20 cm (Fig. 5b).

Gross primary production and net ecosystem exchange

Abiotic variables explained 36% and 33% of variation

in measured GPP and NEE fluxes in the linear

mixed effects model, respectively (Table S3). Both

GPP and NEE were related to WT, T2 and PPFD and

had significant Treatment x PPFD interactions. The

interaction term indicated that the lowered treatment

had lower sensitivity to increasing light levels

(Table S3). The nonlinear model showed that both

GPPpeak and NEEpeak had their greatest magnitude

(i.e., greatest productivity and net C uptake) when

the water table was level with the peat surface

(Table 1 and Fig. 6a, b). Water table position also

had a significant influence on maximum GPPPeak and

NEEPeak both among treatments and between wet

and dry years (Fig. 3c, d). However, reduced magni-

tude of NEEPeak and GPPPeak in the lowered water

table plot during wet years (Fig. 3c, d) was not due

to water table position, as there were no differences

among plots in water table position during wet years

(Fig. 1). This treatment effect on NEEPeak in the low-

ered water table plot occurred in both early (2007–2009) and late (2012–2013) wet years of the study

(Fig. 7). Light response curves for each plot similarly

showed that magnitude of GPPmax was reduced

under drier conditions, but that the lowered water

table treatment further had reduced magnitude of

GPPmax compared to the other plots also during wet

years despite no difference in water table position

(Fig. 8).

0.00

0.02

0.04

0.06

0.08

0 10 20 30 40

R2

of c

orre

latio

nm

odel

resi

dual

v.s

. ΔW

T

Time-lag considered (days)

Methane fluxEcosystem respiration

Fig. 4 Lag effect of changing water table for methane emissions

and ecosystem respiration, considered over time periods

between 1 and 40 days prior to flux measurement. Lag effect

strength was assessed using the coefficient of determination

(R2) between shifts in water table over a period prior to flux

measurements and the residuals of the most parsimonious lin-

ear models for FCH4 and ER (Table S1). The correlations consis-

tently indicated that the linear mixed effects model

overestimated FCH4 when the water table was rising and con-

versely underestimated FCH4 when the water table was falling.

The influence of water table shifts was minor for ER.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2434 D. OLEFELDT et al.

Discussion

Rich fens are a common peatland type in the boreal

biome (Vitt et al., 2000; Whitcomb et al., 2009), but tend

to be understudied compared to boreal poor fens and

bogs because of the complexity of these systems and

the difficulty of making measurements in systems that

flood and dry regularly (Lund et al., 2009; Turetsky

et al., 2014). The experimental design and nine-year

duration of the study allowed the exploration of long-

and short-term effects of water table fluctuations.

Results discussed below further show that the studied

rich fen has both similarities and differences to the

likely impacts of altered hydrological regimes of other

boreal peatland types. Understanding impacts of

altered hydrological regimes, and potential differences

among peatland types, is critical for making predictions

about the future greenhouse gas balance of boreal peat-

lands under future climates.

Long-term effects of experimentally modified droughtseverity

The effects of our wetting and drying treatment on veg-

etation composition were assessed in 2010, six years

after the establishment of the water table manipulations

(Churchill et al., 2015). In brief, the study found that

water table treatment led to no significant changes in

total biomass or vascular net primary productivity, but

that the raised water table plot had slightly increased

abundance of sedges, while the lowered plot had

reduced brown moss cover and increased total vascular

ER = –0.53xNEE + 1.4R² = 0.52 P < 0.01

1.5

2

2.5

3

3.5

4

4.5

–5 –4 –3 –2 –1 0 1

Eco

syst

em re

spira

tion

(μm

ol C

O2

m–2

s–1

)

Net ecosystem exchange(μmol CO2 m–2 s–1)

Control (Wet) Control (Dry)Lowered (Wet) Lowered (Dry)Raised (Wet) Raised (Dry)

–0.3

0

0.3

0.6

0.9

1.2

1.5

1.8

–45 –30 –15 0 15

–NE

E/E

R

Water table position (cm)

(a) (b)

Fig. 5 Relationships between (a) average measured peak growing season (13 June–22 August) ecosystem respiration and net ecosystem

exchange fluxes and (b) the average ratio of measured peak growing season net ecosystem exchange to ecosystem respiration magni-

tude and water table position. Each symbol represents an individual subplot, and error bars represent �1 standard error. Only fluxes

measured under high light conditions (PPFD > 400 lmol m�2 s�1) were included in analysis. Positive water table position indicates

water table above the peat surface. Fitted line shows a significant linear relationship between subplot NEEPeak and ERPeak during wet

years across all treatments. Dry years were 2006, 2010 and 2011, and wet years were 2005, 2007, 2008, 2009, 2012 and 2013.

–10

–8

–6

–4

–2

0

2

4

–60 –40 –20 0 20

Net

eco

syst

em e

xcha

nge

(μm

olC

O2

m–2

s–1

)

Water table position (cm)

ControlLoweredRaised

–9

–12

–15

–6

–3

0

3

–60 –40 –20 0 20Gro

ss –

15–1

2prim

ary

prod

uctio

n(μ

mol

CO

2 m

–2 s

–1)

Water table (cm)

ControlLoweredRaised

(a) (b)

Fig. 6 Relationship between peak growing season (13 June–22 August) net ecosystem exchange and water table position. Only mea-

surements taken under high light conditions (PPFD > 400 lmol m�2 s�1) are included. The dashed line represents a nonlinear regres-

sion based on data from all subplots using Eqn (2), which is described in Table 1. Positive water table position indicates water table

above the peat surface.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2435

green area in comparison with the control plot. The

abundance of drought-tolerant shrub species did not

increase in the lowered water table treatment, which

has been observed in long-term water table manipula-

tions in nutrient-poor peatlands (Laine et al., 1996;

Weltzin et al., 2000). This is likely due to how our mea-

surement period included years when the water table

remained above the peat surface throughout the grow-

ing season even in the lowered water table plot. Hence,

in the lowered water table plot, exacerbated drought

conditions during dry years caused the loss of drought-

sensitive brown mosses but wet years still likely pre-

vented the establishment of drought-tolerant but

flood-sensitive shrubs. This highlights how natural

hydrological variability of rich fens influences stability of

vegetation communities, which in turn has implications

for impacts of hydrological variability on C cycling.

The lowered water table treatment exhibited altered

relationships between CO2 fluxes and abiotic variables

relative to the other treatment plots, likely linked to

shifts in vegetation composition. Under conditions

optimal for maximum GPPpeak and ERpeak, i.e., mea-

surements taken under full sunlight when the water

table was near the peat surface, the lowered treatment

plot had GPP and ER fluxes that were reduced in

magnitude by ~15% compared with control and

raised water table plots. Reduced GPP is likely to

have led to lower rates of autotrophic respiration

(Han et al., 2014) and thus caused the observed reduc-

tion in ER. The net effect was reduced magnitude of

NEE, indicating reduced capacity for C uptake during

summers. These treatment effects on the lowered

water table plot were evident in the lowered plot

after just 2 years of water table manipulation (Chivers

et al., 2009; Churchill et al., 2015), and this study

shows that these effects have been maintained over

the 9 years of the experiment.

Water table variability and CO2 fluxes

Our chamber flux measurements corroborate results

from an eddy covariance study carried out at the site

(Euskirchen et al., 2014), showing that both NEE and

GPP peak (i.e., have maximum rates of photosynthesis

and C uptake) when the water table is approximately

level with the peat surface. The optimal water table

position at the APEX fen for NEE (�2 to +5 cm) is wet-

ter than has been observed in a boreal poor fen (�10 to

�20 cm) (Yurova et al., 2007), likely due to differences

in vegetation composition and rooting depths, where

the APEX fen has relatively more emergent vascular

plants and less Sphagnum mosses than the poor fen.

Reduced photosynthetic uptake of mosses is likely

when the water table is above the peat surface. Con-

versely, reduced uptake during drier periods could be

due to plant moisture stress – particularly for bryo-

phytes (Turetsky, 2003) and dwarf shrubs (Lindroth

et al., 2007).

Relative to relationships with NEE and GPP, the

influence of water table position on ER was weaker but

indicated two optima: when the water table was just

under the peat surface and again once the water table

dropped below �40 cm. Previous peatland studies

have found conflicting relationships (including posi-

tive, negative and no relationships) between ER and

water table position (Chimner & Cooper, 2003; Lafleur

et al., 2005; Juszczak et al., 2013; McConnell et al., 2013;

Ballantyne et al., 2014). These conflicting results may

partially be due to previous studies not being able, as

in this study, to determine the influence of water table

position over a wide water table range under compara-

ble soil temperatures. The influence of water table posi-

tion on ER is also likely obscured by the fact that ER is

the sum of autotrophic and heterotrophic respiration

that each is affected independently by water table posi-

tion. Peat mineralization rates (heterotrophic respira-

tion) are expected to decrease with wetter conditions as

incubation experiments show rates reduced by on aver-

age by 80% under anaerobic conditions compared to

aerobic conditions (Schuur et al., 2015). However, pho-

tosynthetically driven respiration, including both strict

–5

–4

–3

–2

–1

0

Early Wet Years(2007-09)

Late Wet Years(2012-13)

Net

eco

syst

em e

xcha

nge

(μm

ol C

O2

m–2

s–1

)

ControlLoweredRaised

Fig. 7 Average peak growing season (13 June–22 August) net

ecosystem exchange fluxes among treatments measured under

high light conditions (PPFD > 400 lmol m�2 s�1) during early

(2007–2009) and late (2012–2013) wet years in the study. Error

bars represent �2 standard errors. Two-way ANOVA followed by

a multiple comparison using a Bonferroni correction indicated

that the lowered treatment had lower magnitude fluxes during

both periods (P < 0.1). In wet years, there was no difference in

water table position among treatments (control, lowered and

raised) (see Fig. 1). Lower magnitude fluxes during the late wet

years are primarily due to measurements being taken later in

the season on average, with average dates July 3 and August 7

for early and late wet years, respectively (see Fig. S3 for season-

ality of NEE).

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2436 D. OLEFELDT et al.

autotrophic respiration (foliage, stems and roots) and

rhizospheric respiration of root exudates, often domi-

nates wetland ER (Frolking et al., 2002; Crow & Wieder,

2005) and can be highly responsive to short-term varia-

tion in GPP (Han et al., 2014). The observed relation-

ship in this study between NEE and ER among

subplots under high light conditions in wet years sug-

gested that photosynthetically driven respiration repre-

sented 35 � 14% of GPP under such conditions. This

implies that photosynthetically driven respiration rep-

resented ~70 � 28% of ER during such periods

(GPPpeak 9 0.35 � 0.14/ERpeak). If photosynthetically

driven respiration as a fraction of GPP can be assumed

similar during dry years, it follows that photosyntheti-

cally driven respiration as a fraction of ER under simi-

lar light conditions in dry years drops to 63 � 25,

50 � 20 and 41 � 16%, respectively, in the raised, con-

trol and lowered treatment subplots, with concurrent

increases in heterotrophic contribution.

Increased importance of heterotrophic respiration

during drier periods was further supported by the

increasing predictive capability of soil temperatures at

25 cm depth for ER during the driest periods. A poten-

tially interesting influence on heterotrophic respiration

is thus that deep soil temperatures in dry years are sub-

stantially colder than in wet years, thus suppressing

rates of peat mineralization despite aerobic conditions

(c.f. Ise et al., 2008).

Lag effects on ER following droughts have been

observed in nutrient-poor peatlands due to temporary

reductions of biogeochemical constraints on anaerobic

microbial activity. In nutrient-poor peatlands, droughts

initiate a biogeochemical cascade where increased aero-

bic microbial activity causes a release of nutrients and

increased pH, which in turn significantly increases

anaerobic rates of peat mineralization following rewet-

ting when compared to before the drought (Fenner &

Freeman, 2011). In this study, we found neither short-

term (1–50 days) nor inter-annual lag effects on ER

linked to shifts in water table position. Furthermore, a

peat incubation experiment using soil organic matter

from the APEX treatments showed only a minor differ-

ence between aerobic and anaerobic rates of peat min-

eralization (Kane et al., 2013), with much higher rates of

anaerobic CO2 production than expected. These results

support the interpretation that biogeochemical con-

straints on anaerobic microbial activity are less strict in

more nutrient rich peatlands with higher pH (Ye et al.,

2012) and that ER in rich fens thus is less likely to exhi-

bit inter-annual lag effects following drought.

Methane emissions and water table variability

It is well established in the literature, and corroborated

in our results, that the balance between anaerobic CH4

production below the water table and oxidation above

–15

–12

–9

–6

–3

0

0 1000 2000–15

–12

–9

–6

–3

0

0 1000 2000–15

–12

–9

–6

–3

0

0 1000 2000

Photosynthetic photon flux density (μmol m–2 s–1)

Gro

ss p

rimar

y pr

oduc

tion

(μm

ol C

O2

m–2

s–1

)(a) Control treatment (b) Lowered treatment (c) Raised treatment

Wet: –8.5 ±2.0, 290 ±180Dry: –8.0 ±2.2, 670 ±420

Wet: –7.2 ±1.8, 440 ±310Dry: –4.4 ±0.8, 120 ±130

Wet: –8.7 ±1.9, 420 ±170Dry: –8.6 ±2.3, 490 ±340

Fig. 8 Relationships between gross primary production and photosynthetic photon flux density among treatments and between dry

and wet years. Comparison uses data from peak growing season only. Parameters estimates � 95% CI are reported for GPPmax (maxi-

mum rate of GPP under light saturation) and k (light level where half of GPPmax is attained) (Eqn 2). Dotted lines are fits (Eqn 2) for

dry years, and continuous lines are for wet years. Water table position varied among treatment plots during dry years, but not during

wet years (Fig. 1).

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2437

it leads to rapidly increasing CH4 fluxes with a higher

water table position (Bridgham et al., 2013). Our results

further showed short-term lag effects on CH4 emissions

due to water table fluctuation. This lag effect was eco-

logically significant as our linear model over- and

underestimated CH4 emissions by ~25% when the

water table had raised or dropped by 5 cm over the

preceding 5 days, respectively. These short-term lag

effects may arise due to physical processes such as

changes in hydrostatic pressure, due to suppression of

methanogens until alternate electron acceptors are

depleted (Knorr and Blodau, 2009; Deppe et al., 2010)

or due to differential growth rates between methano-

gens and methanotrophs (Segers, 1998). Given that

water table position generally drops over the season,

our results suggest that short-term lag effects are

required to be taken into account when modeling

methane emissions from northern peatlands.

Methane emissions were strongly associated with

soil temperatures at 25 cm depth. Average peak

growing season soil temperature at 25 cm over the

9 years was uncorrelated with air temperatures, but

was on average ~3.5 °C warmer in wet than in dry

years. Higher deep soil temperatures in wet years are

likely due to increased soil thermal conductivity of

flooded soils. Our empirical models indicate that a

3.5 °C increase at 25 cm soil depth when the water

table is level with the peat surface leads to 85% to

120% increases in methane emissions. As such, higher

methane emissions in wet years are indicated to be

equally due to increased rates of methanogenesis in

warmer anaerobic peat layers as it due to reduced

capacity for methanotrophy in a thinner aerobic layer.

A coupled hydrological and biogeochemical model of

wetland greenhouse gases has accordingly shown that

wet years can lead to increased soil temperatures,

which in turn raise CH4 emissions (Grant, 2015). This

connection between water table position and deep soil

temperatures is thus important to consider not to

underestimate future methane emissions in wetlands

that occasionally flood.

Climate change implications for boreal rich fens

Studies of water table and soil temperature influences

on methane emissions and the balance between GPP

and ER for the overall C balance of northern peat-

lands have shown that different peatland types can

be expected to respond differently (Bubier et al., 1998;

Sulman et al., 2010; Turetsky et al., 2014). Our results

show that altered frequency and severity of droughts

and floods will have a strong influence on the overall

C balance of rich fens like APEX. Eddy covariance

measurements have found the site to be a significant

C sink (~80 g C m�2 yr�1) during wet years (2012

and 2013), but the record does not yet include a dry

year for comparison (Euskirchen et al., 2014). Our

results show that dry years lead to reduced capacity

for C uptake as a result of inhibited GPP, while ER

magnitude is sustained by a shift in dominance from

autotrophic to heterotrophic respiration. Similar nega-

tive influence of dry years on photosynthetic capacity

has been suggested based on eddy covariance mea-

surements for a boreal rich fen in Finland (Aurela

et al., 2009). However, we further observed reduced

rates of peak growing season GPP and NEE in the

lowered treatment during subsequent wet years

despite there being no difference in water table posi-

tion among treatments in wet years. This indicates

that extreme droughts have long-term, inter-annual,

effects on C uptake in rich fens, likely due to reduced

photosynthetic capacity of an altered vegetation com-

munity.

It is not certain that climate change will lead to

increased frequency or severity of summer droughts in

rich fens as is expected for boreal nutrient-poor peat-

lands (Wu & Roulet, 2014). Given the amplification of

the hydrological cycle, wetness of rich fens may not

respond to climate like the overall landscape given the

importance surface water and groundwater inflows

(Walvoord & Striegl, 2007; Olefeldt & Roulet, 2012). For

example, the APEX rich fen is located on a floodplain

and remained flooded throughout the 2013 growing

season despite less than average seasonal precipitation.

There is currently a limited understanding of how cli-

mate change may affect regional hydrology and thus

hydrological regimes of rich fens. Projecting future

hydrological regimes of rich fens thus represents a key

uncertainty future greenhouse gas exchange of north-

ern peatlands overall.

Acknowledgements

This APEX has been supported by National Science Founda-tion grants (DEB-0425328, DEB-0724514 and DEB-0830997) toM.R.T, A.D.M, J.H., E.E. and E.S.K., the Bonanza Creek Long-Term Ecological Research program (funded jointly by NSFGrant DEB-0620579, an USDA Forest Service Pacific NorthwestResearch Grant PNW01-JV11261952-231) and U.S. GeologicalSurvey Climate and Land Use Change Program and ClimateScience Center grant funds to J.H., A.D.M., M.W. and E.E.During manuscript compilation and writing, D.O. was sup-ported by a Campus Alberta Innovates Program grant. Anyuse of trade, firm or product names is for descriptive pur-poses only and does not imply endorsement by the U.S.Government. We thank all collaborators and students whohave contributed over the years to the APEX project, namelyBill Cable, Colin Edgar, Michael Waddington, Jamie Hollings-worth, Teresa Hollingsworth, Rebecca Finger, Amy Churchill,Nicole McConnell and Molly Chivers. The authors claim noconflicts of interest.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2438 D. OLEFELDT et al.

References

Adkinson AC, Syed KH, Flanagan LB (2011) Contrasting responses of growing season

ecosystem CO2 exchange to variation in temperature and water table depth in two

peatlands in northern Alberta, Canada. Journal of Geophysical Research, 116, G01004.

Aurela M, Lohila A, Tuovinen J-P, Hatakka J, Riutta T, Laurila T (2009) Carbon diox-

ide exchange of a northern boreal fen. Boreal Environmental Research, 14, 699–710.

Ballantyne DM, Hribljan JA, Pypker TG, Chimner RA (2014) Long-term water table

manipulations alter peatland gaseous carbon fluxes in Northern Michigan. Wet-

lands Ecology and Management, 22, 35–47.

Batjes NH (1996) Total carbon and nitrogen in the soils of the world. European Journal

of Soil Science, 47, 151–163.

Berg EE, McDonnel Hillman K, Dial R, DeRuwe A (2009) Recent woody invasion of

wetlands on the Kenai Peninsula Lowlands, south-central Alaska: a major regime

shift after 18 000 years of wet Sphagnum-sedge peat recruitment. Canadian Journal

of Forest Research, 39, 2033–2046.

Bergamschi P, Frankenberg C, Meirink JF et al. (2007) Satellite chartography of atmo-

spheric methane from SCIAMACHY on board ENVISAT: 2. Evaluation based on

inverse model simulations. Journal of Geophysical Research - Atmospheres, 112,

D02304.

Bridgham SD, Cadillo-Quiroz H, Keller JK, Zhuang Q (2013) Methane emissions from

wetlands: biogeochemical, microbial, and modeling perspectives from local to glo-

bal scales. Global Change Biology, 19, 1325–1346.

Bubier JL, Crill PM, Moore TR, Savage K, Varner RK (1998) Seasonal patterns and

controls on net ecosystem CO2 exchange in a boreal peatland complex. Global Bio-

geochemical Cycles, 12, 703–714.

Carrol P, Crill P (1997) Carbon balance of a temperate poor fen. Global Biogeochemical

Cycles, 11, 349–356.

Chimner RA, Cooper DJ (2003) Influence of water table levels on CO2 emissions in a

Colorado subalpine fen: an in situ microcosm study. Soil Biology and Biogeochem-

istry, 35, 345–351.

Chivers MR, Turetsky MR, Waddington JM, Harden JW, McGuire AD (2009) Effects

of experimental water table and temperature manipulations on ecosystem CO2

fluxes in an Alaskan rich fen. Ecosystems, 12, 1329–1342.

Churchill AC, Turetsky MR, McGuire AD, Hollingsworth TN (2015) Response of

plant community structure and primary productivity to experimental drought and

flooding in an Alaskan fen. Canadian Journal of Forest Research, 45, 185–193.

Clymo RS, Turunen J, Tolonen K (1998) Carbon accumulation in peatland. Oikos, 81,

368–388.

Crow SE, Wieder KR (2005) Sources of CO2 emissions from a northern peatland: root

respiration, exudation, and decomposition. Ecology, 86, 1825–1834.

Deppe M, McKnight D, Blodau C (2010) Effects of short-term drying and irrigation

on electron flow in mesocosms of a northern bog and an alpine fen. Environmental

Science and Technology, 44, 80–86.

Dowrick DJ, Freeman C, Lock MA, Reynolds B (2006) Sulphate reduction and the

suppression of peatland methane emissions following summer drought. Geoderma,

132, 384–390.

Euskirchen ES, Edgar CW, Turetsky MR, Waldrop MP, Harden JW (2014) Differential

response of carbon fluxes to climate in three peatland ecosystems that vary in the

presence and stability of permafrost. Journal of Geophysical Research, 119, 1576–1595.

Fenner N, Freeman C (2011) Drought-induced carbon loss in peatlands. Nature Geo-

science, 4, 895–900.

Frolking S, Roulet NT (2007) Holocene radiate forcing impact of northern peatland

carbon accumulation and methane emissions. Global Change Biology, 13, 1079–1088.

Frolking S, Roulet NT, Moore TR, Lafleur PM, Bubier JL, Crill PM (2002) Modeling

seasonal to annual carbon balance of Mer Bleue Bog, Ontario, Canada. Global Bio-

geochemical Cycles, 16. doi:10.1029/2001GB001457.

Grant RF (2015) Ecosystem CO2 and CH4 exchange in a mixed tundra and a fen

within a hydrologically diverse Arctic landscape: 2. Modeled impacts of climate

change. Journal of Geophysical Research Biogeosciences, 120, 1388–1406.

Han G, Luo Y, Li D, Xia J, Xing Q, Yu J (2014) Ecosystem photosynthesis regulates soil

respiration on a diurnal scale with a short-term time lag in a coastal wetland. Soil

Biology and Biochemistry, 68, 85–94.

Hartmann DL, Klein Tank AMG, Rusticucci M et al. (2013) Observations: atmosphere

and surface. In: Climate Change 2013: The Physical Science Basis. Contribution of Work-

ing Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate

Change (eds Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J,

Nauels A, Xia Y, Bex V, Midgley PM), pp. 159–254. Cambridge University Press,

Cambridge, UK and New York, NY, USA.

Hinzman LD, Bettez ND, Bolton WR et al. (2005) Evidence and implications of recent cli-

mate change in northern Alaska and other arctic regions. Climatic Change, 72, 251–298.

Hinzman LD, Viereck LA, Adams PC, Romanovsky VE, Yoshikawa K (2006) Climatic

and permafrost dynamics of the Alaskan boreal forest. In: Alaska’s Changing Boreal

Forest (eds Chapin FS III, Oswood MW, Van Cleve K, Viereck LA, Verbyla DL),

pp. 39–61. Oxford University Press, New York, NY, USA.

Ise T, Dunn AL, Wofsy SC, Moorcroft PR (2008) High sensitivity of peat decom-

position to climate change through water-table feedback. Nature Geoscience, 1,

763–766.

Juszczak R, Humphreys E, Acosta M, Michalak-Galczewska M, Kayzer D, Olejnik J

(2013) Ecosystem respiration in a heterogeneous temperate peatland and its sensi-

tivity to peat temperature and water table depth. Plant and Soil, 366, 505–520.

Kane ES, Chivers MR, Turetsky MR et al. (2013) Response of anaerobic carbon cycling

to water table manipulations in an Alaskan rich fen. Soil Biology and Biochemistry,

58, 50–60.

Knorr K-H, Blodau C (2009) Impact of experimental drought and rewetting on redox

transformations and methanogenesis in mesocosms of a northern fen soil. Soil Biol-

ogy and Biochemistry, 41, 1187–1198.

Lafleur PM, Moore TR, Roulet NT, Frolking S (2005) Ecosystem respiration in a cool

temperate bog depends on peat temperature but not water table. Ecosystems, 8,

619–629.

Laiho R (2006) Decomposition in peatlands: reconciling seemingly contrasting

results on the impacts of lowered water levels. Soil Biology and Biochemistry, 38,

2011–2024.

Laine J, Silvola J, Tolonen K et al. (1996) Effect of water-level drawdown on global cli-

matic warming: northern peatlands. Ambio, 25, 179–184.

Limpens J, Berendse F, Blodau C et al. (2008) Peatlands and the carbon cycle: from

local processes to global implications – a synthesis. Biogeosciences, 5, 1475–1491.

Lindroth A, Lund M, Nilsson M et al. (2007) Environmental controls on the CO2

exchange in north European mires. Tellus, 59B, 812–825.

Lund M, Lafleur PM, Roulet NT et al. (2009) Variability in exchange of CO2

across 12 northern peatland and tundra sites. Global Change Biology, 16, 2436–

2448.

Lund M, Christensen TR, Lindroth A, Schubert P (2012) Effects of drought conditions

on the carbon dioxide dynamics in a temperate peatland. Environmental Research

Letters, 7, 045704.

McConnell NA, Turetsky MR, McGuire AD, Kane ES, Waldrop MP, Harden JW

(2013) Controls on ecosystem and root respiration across a permafrost and wet-

land gradient in interior Alaska. Environmental Research Letters, 8, 045029.

Moore TR, Knowles R (1989) The influence of water table levels on methane and carbon

dioxide emissions from peatland soils. Canadian Journal of Soil Science, 69, 33–38.

Olefeldt D, Roulet N (2012) Effects of permafrost and hydrology on the composition

and transport of dissolved organic carbon in a subarctic peatland complex. Journal

of Geophysical Research, 117, G01005.

Olefeldt D, Turetsky MR, Crill PM, McGuire AD (2013) Environmental and physical

controls on northern terrestrial methane emissions across permafrost zones. Global

Change Biology, 19, 589–603.

Peichl M, Sonnentag O, Nilsson M (2015) Bringing color into the picture: using digital

repeat photography to investigate phenology controls of the carbon dioxide

exchange in a boreal mire. Ecosystems, 18, 115–131.

Rawlins MA, Steele M, Holland MM et al. (2010) Analysis of the arctic system for

freshwater cycle intensification: observations and expectations. Journal of Climate,

23, 5715–5737.

Roulet NT, Lafleur PM, Richard PJ, Moore TR, Humphreys ER, Bubier J (2007) Con-

temporary carbon balance and late Holocene carbon accumulation in a northern

peatland. Global Change Biology, 13, 397–411.

Rouse WR (1998) A water balance model for a subarctic sedge fen and its application

to climate change. Climatic Change, 38, 207–234.

Schuur EAG, McGuire AD, Sch€adel C et al. (2015) Climate change and the permafrost

carbon feedback. Nature, 520, 171–179.

Segers R (1998) Methane production and methane consumption: a review of pro-

cesses underlying wetland methane fluxes. Biogeochemistry, 41, 23–51.

Serreze MC, Francis JA (2006) The Arctic amplification debate. Climatic Change, 76,

241–261.

Silvola J, Alm J, Ahlholm U, Nyk€anen H, Martikainen PJ (1996) CO2 fluxes from peat

in boreal mires under varying temperature and moisture conditions. Journal of

Ecology, 84, 219–228.

Sulman BN, Desai AR, Cook BD, Saliendra N, Mackay DS (2009) Contrasting carbon

dioxide fluxes between a drying shrub wetland in Northern Wisconsin, USA, and

nearby forests. Biogeosciences, 6, 1115–1126.

Sulman BN, Desai AR, Saliendra NZ et al. (2010) CO2 fluxes at northern fens and bogs

have opposite responses to inter-annual fluctuations in water table. Geophysical

Research Letters, 37, L19702.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

WATER TABLE CONTROL ON FEN GHG EXCHANGE 2439

Sun CL, Brauer SL, Cadillo-Quiroz H, Zinder SH, Yavitt JB (2012) Seasonal changes

in methanogenesis and methanogenic community in three peatlands, New York

State. Frontiers in Microbiology, 3, 81.

Tardif S, St-Hilaire A, Roy R, Payette S (2015) Water budget analysis of small forested

boreal watersheds: comparison of Sphagnum bog, patterned fen and lake domi-

nated downstream areas in the La Grande River Region, Quebec. Hydrology

Research, 46, 106–120.

Tuitilla E-S, Vasander H, Laine J (2004) Sensitivity of C sequestration in reintroduced

sphagnum to water-level variation in a cutaway peatland. Restoration Ecology, 12,

483–493.

Turetsky MR (2003) The role of bryophytes in carbon and nitrogen cycling. The Bryol-

ogist, 106, 395–409.

Turetsky MR, Treat CC, Waldrop MP, Waddington JM, Harden JW, McGuire AD

(2008) Short-term response of methane fluxes and methanogen activity to water

table and soil warming manipulations in an Alaskan peatland. Journal of Geophysi-

cal Research, 113, G00A10.

Turetsky MR, Kotowska A, Bubier J et al. (2014) A synthesis of methane emissions from

71 northern, temperate, and subtropical wetlands. Global Change Biology, 20, 2183–2197.

Vitt DH, Halsey LA, Bauer IE, Campbell C (2000) Spatial and temporal trends in car-

bon storage of peatlands of continental western Canada through the Holocene.

Canadian Journal of Earth Sciences, 37, 683–693.

Walvoord MA, Striegl RG (2007) Increased groundwater to stream discharge from

permafrost thawing in the Yukon River basin: potential impacts on lateral export

of carbon and nitrogen. Geophysical Research Letters, 34, L12402.

Ward SE, Ostle NJ, Oakley S, Quirk H, Henrys PA, Bardgett RD (2013) Warming

effects on greenhouse gas fluxes in peatlands are modulated by vegetation compo-

sition. Ecology Letters, 16, 1285–1293.

Weltzin JF, Pastor J, Harth C, Bridgham SD, Updegraff K, Chapin C (2000) Respone

of bog and fen plant communities to warming and water table manipulations.

Ecology, 81, 3464–3478.

Wendler G, Shulski M (2009) A century of climate change for Fairbanks, Alaska. Arc-

tic, 62, 295–300.

Whitcomb J, Moghaddam M, McDonald K, Kellndorfer J, Podest E (2009) Mapping

vegetated wetlands of Alaska using L-band radar satellite imagery. Canadian Jour-

nal of Remote Sensing, 35, 54–72.

Wu J, Roulet NT (2014) Climate change reduces the capacity of northern peatlands to

absorb the atmospheric carbon dioxide: the different responses of bogs and fens.

Global Biogeochemical Cycles, 28, 1005–1024.

Ye R, Jin Q, Bohannan B, Keller JK, McAllister SA, Bridgham SD (2012) pH controls

over anaerobic carbon mineralization, the efficiency of methane production, and

methanogenic pathways in peatlands across an ombrotrophic–minerotrophic gra-

dient. Soil Biology and Biochemistry, 54, 36–47.

Yu Z (2012) Northern peatland carbon stocks and dynamics: a review. Biogeosciences,

9, 4071–4085.

Yurova A, Wolf A, Sagerfors J, Nilsson M (2007) Variations in net ecosys-

tem exchange of carbon dioxide in a boreal mire: modeling mechanisms

linked to water table position. Journal of Geophysical Research Biogeosciences,

112, G02025.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Figure S1. Relationships between (a) average daily soil temperature at 25 cm depth and soil temperatures at 25 cm in the raisedand lowered water table treatment, and (b) average daily soil temperature in subplots without open top chambers (OTC) and sub-plots with OTC.Figure S2. Median measured daytime fluxes of methane fluxes and ecosystem respiration throughout the season.Figure S3. Median measured fluxes under high light conditions (PPFD > 400 lmol m�2 s�1) throughout the season of (a) NEE and(b) GPP. Dry years were 2006, 2010 and 2011, while wet years were 2005, 2007, 2008, 2009, 2012, and 2013.Figure S4. Average peak growing season (DOY 165-234) soil temperatures (2005–2013) at 2 and 25 cm peat depths plotted against(a) average peak growing season air temperatures, and (b) average peak growing season water table position in the control plot.Table S1. Results from linear mixed effects model analyzing controls on measured daytime ER and Log10 transformed CH4 fluxesacross three water table treatments: ANOVA of the marginal effects of the parameters and the final regression model with estimates ofsignificant fixed effects coefficients.Table S2. Estimated parameters for models examining the non-linear temperature dependencies of methane fluxes and ecosystemrespiration (Eq. 1: Flux = A 9 QT=10

A ).Table S3. Results from linear mixed effects model analyzing controls on measured GPP and NEE fluxes across three water tabletreatments: ANOVA of the marginal effects of the parameters and the final regression model with estimates of significant fixed effectscoefficients.

© 2017 John Wiley & Sons Ltd, Global Change Biology, 23, 2428–2440

2440 D. OLEFELDT et al.