Introduction
Northern peatlands cover less than 3 % of the Earth's land surface but
have sequestered approximately one-third of the global store of terrestrial
carbon (Gorham, 1991; Yu, 2012), serving an important role in the global
carbon cycle (Roulet et al., 2007; Limpens et al., 2008). However, this
accumulated carbon is expected to be severely affected by climate change due
to changes in temperature and precipitation patterns (Kettles and Tarnocai,
1999; Li et al., 2007; Gallego-Sala and Prentice, 2013). The net uptake of
carbon in peatlands is due to relatively cool and moist conditions that
promote the slower release of CO2 by ecosystem respiration (ER) than uptake
by gross ecosystem productivity (GEP) (Frolking et al., 2002; Rydin and
Jeglum, 2006; Frolking et al., 2010). Maintenance of the carbon sink
function of peatlands is dependent on a high water table (WT) to minimize
respiration (Alm et al., 1999; Strack et al., 2006; Riutta et al., 2007) and
high near-surface soil moisture to enable high rates of Sphagnum moss photosynthesis
(Waddington and Roulet, 2000; Moore et al., 2002; Adkinson and Humphreys,
2011), both of which are controlled by the precipitation regime. Changing
climate and weather patterns are leading to an intensification of the
precipitation regime (Easterling et al., 2000; Cao and Ma, 2009; Diffenbaugh
and Field, 2013), with increasingly longer periods without rain that
threaten the sink function of northern peatlands (Nijp et al., 2017).
Climate projections for boreal and temperate regions of the Northern
Hemisphere suggest larger but less frequent precipitation events, with
limited to no increase in absolute seasonal and annual precipitation
(Trenberth, 2011; IPCC, 2013; Sillmann et al., 2013; Wang et al., 2014;
Westra et al., 2014). This redistribution of the precipitation regime,
especially during the growing season, is expected to lead to decreases in
near-surface soil moisture, increased soil moisture variability, and deeper
WTs(Knapp et al., 2002; Gerten et al., 2008; Vervoort and van der Zee, 2008;
Wu et al., 2012). Lower WT levels and near-surface soil moisture due to seasonal
drought or temperature-induced increased evapotranspiration (ET) have been shown to lead to
reduced net carbon uptake and/or greater rates of CO2 emission from
peatlands (Carroll and Crill, 1997; Alm et al., 1999; Tuittila et al., 2004;
Strack et al., 2006; Riutta et al., 2007). Additionally, changes to peatland
hydrology through drought have led to changes in vegetation biomass and
community composition, which in turn can alter peatland CO2 exchange
depending on the photosynthesis to respiration ratio of the newly
established communities (Buttler et al, 2015; Potvin et al., 2015; Churchill
et al., 2015; Dieleman et al., 2015). However, these experimental and
monitoring studies typically reduce total precipitation input during the
study period, while much less research has focussed on the redistribution of
rainfall whilst maintaining precipitation totals.
Most studies examining the impact of precipitation frequency on ecosystems
have focussed on grassland systems (Hoover et al., 2014; Knapp et al., 2015;
Wilcox et al., 2015; Didiano et al., 2016), with comparatively fewer studies
in wetland environments. Riparian marsh species' biomass accumulation is
negatively affected by month-long periods without added precipitation
(Garssen et al., 2014). In peatland systems, shifts in rainfall frequency
have been shown to affect net primary production from Sphagnum moss (Robroek et al.,
2009; Nijp et al., 2014) and methane emission (Radu and Duval, 2018a).
Sphagnum moss photosynthesis, in particular, responds quickly to trace amounts of
precipitation input (Strack and Price, 2009; Adkinson and Humphries, 2011);
however, precipitation inputs are only available to these peat-building
species for 2–3 days before it is evaporated (Ketcheson and Price, 2014). In
the absence of precipitation, Sphagnum depends on capillary rise from the saturated
zone to the photosynthesizing capitula (Clymo and Hayward, 1982). When WTs
are too low, and precipitation is absent, soil water tensions increase and
hyaline cells drain, causing desiccation and reduced photosynthesis
(Thompson and Waddington, 2008; Strack et al., 2009; McCarter and Price,
2014).
In addition to the non-vascular Sphagna species, peatlands can be dominated by
shrub and graminoid plant functional types, primarily ericaceous shrubs and
sedges, respectively (Rydin and Jeglum, 2006). Vascular plants
photosynthesize as long as the component cells retain turgor pressure, and
are less susceptible to periods of low precipitation and WT drawdowns
(Malmer et al., 1994; Vile et al., 2011). Peatland shrubs typically have
increased productivity when WTs are lowered (Weltzin et al., 2001; Murphy et
al., 2009; Bragazza et al., 2013; Munir et al., 2015). Sedges have deep roots (40–100 cm below
peat surface) aided by aerenchyma that transport oxygen to lower depths
(Silvan et al., 2004); thus, they are tolerant of high WT conditions, but have
also been documented to perform well under low WT conditions (Fenner et al.,
2007; Dieleman et al., 2015). Additionally, plant and soil respiration rates
differ between moss, sedge, and shrub communities (Chimner, 2004; Juszczak
et al., 2012; Duval and Radu, 2018). Since vascular plants can comprise a
significant portion of peatland productivity and respiration (Szumigalski
and Bayley, 1996; Moore et al., 2002; Riutta et al., 2007; Korrensalo et al.,
2017), their responses to climate change in concert with Sphagnum must be taken into
account when assessing peatland carbon cycling.
Previous studies of the effect of lowered water tables on different plant
functional types to study the impacts of climate change on peatland carbon
cycling have not included the importance of precipitation frequency (Riutta
et al., 2007; Churchill et al., 2015; Potvin et al., 2015). Recent research
on the interaction between peatland WT position, carbon cycling, and
precipitation frequency has focussed on Sphagnum moss (Nijp et al., 2014, 2015, 2017).
We have recently extended on this research by
showing that decreasing rainfall frequency increased vascular plant cover in
a cool temperate poor fen in southern Ontario. This consequently led to contracting
responses of the seasonal CO2 balance between community types, with
greater flux of CO2 to the atmosphere from the mosses but greater
CO2 sequestration in sedge and shrub communities (Radu and Duval,
2018b). However, natural variability and logistical constraints in the field
prevented us from making a formal assessment of the hydrological controls on the
observed differences in NEE. A research gap exists at the intersection
of precipitation frequency and its effect on peatland hydrology and CO2
for a variety of peatland community types. Therefore, the objectives of this
study are as follows: (i) to investigate the effect of changing precipitation frequency
on peatland hydrology, comprising WT position, VMC, soil water tension, and
evapotranspiration (ET), in three common peatland vegetation communities –
Sphagnum moss only, sedge with Sphagnum, and Sphagnum
with ericaceous shrubs; and (ii) to determine the
relationship between hydrologic conditions under the different rainfall
frequency treatments and GEP, ER, and net ecosystem CO2 exchange of
those communities. We altered precipitation frequency without changing total
precipitation amount in both in situ field plots and lab monoliths through
commensurate changes to event magnitudes.
Methods
Study site
The study was carried out on peat monoliths collected from an undisturbed
poor fen in southern Ontario, Canada (44∘15′13.34′′ N,
80∘20′46.83′′ W). Vegetation was dominated by Sphagnum moss (particularly
S. capillifolium but also S. rubellum, S. fuscum, and
S. magellanicum), sedges (Carex oligosperma and Eriophorum vaginatum), and shrubs (mostly Chamaedaphne calyculata, as well as
Rhododendron groenlandicum and Vaccinium uliginosum). Moss
ground cover is close to 100 % throughout the site, except in
areas with mature shrubs, where ground cover averaged 15 %. Average peat
depth in the sample area is 2.1 m overlaying a sandy silt till substrate
(Burwasser, 1974). The climate near the site is characterized by a mean
annual temperature of 6.4∘ C and a mean annual precipitation of
996 mm (1981–2010 normal at Ruskview, ON station, data available:
http://climate.weather.gc.ca/climate_normals/, last access: 22 June 2018). Rainfall events
>0.2 mm occur on 43 % of the days during the early May –
end of September growing season at this climate station. Over the growing
season of 2015 at this site we manipulated the rainfall frequency whilst holding
total seasonal rain constant; this was undertaken by utilizing rainout shelters over areas dominated
by moss, sedge, and shrub communities. We measured plant community changes
and CO2 exchange in response to these manipulated rainfall regimes at
this site. Details of this field setup and experiment can be found in Radu
and Duval (2018b). Here we focus on the hydrological process controls on the
effect of rainfall frequency on CO2 dynamics through an experiment
under controlled laboratory conditions. We also analyzed the field data in
the same manner as the laboratory study.
Experimental setup
Intact peat cores (30 cm diameter × 40 cm height) were collected from the
peatland to investigate how precipitation frequency affects CO2
exchange under controlled climate and WT regimes. Details of core collection
and setup can be found in Radu and Duval (2018a). Briefly, cores of each of
the three vegetation communities were placed in an environment-controlled
chamber (FXC-19 Chamber, BioChambers, Winnipeg, Manitoba, Canada) and water
tables were kept at -5 cm for an acclimatization period to the chamber
conditions. Climate conditions of the chamber can be found in Table S1 in the Supplement.
We manipulated rainfall frequency over a 4-month study period. Three
precipitation frequency treatments were randomly assigned to monoliths of
each vegetation community type: 3 events/week (“HiFreq-lab”), 1 event/week
(“MedFreq-lab”), and 1 event/2 weeks (“LowFreq-lab”). Simulated rainwater
was a diluted Rudolph nutrient solution (Rudolph et al., 1988; Faubert and
Rochefort, 2002) to limit detrimental effects to Sphagnum moss growth (Dieleman et
al., 2015). The amount of water added for the individual events was adjusted
such that the total amount of water was the same between treatments for each
2-week period (see Table S2 for treatment details). At the beginning of
the study period, all water levels were set to -5 cm (“high”) and after
2 months were adjusted to -15 cm (“low”). Within each of these two 2-month
periods, WT positions were allowed to naturally fluctuate with the addition
of rainwater and the loss of water through ET to simulate field conditions.
Perforated PVC wells (1.27 cm diameter) covered with 250 µm Nitex
mesh were carefully inserted into peat cores to monitor WT levels;
measurements were made three times weekly in all monoliths. VMC was measured
with EC-5 soil moisture sensors installed vertically into the peat of each
monolith for an integrated depth of 0–5 cm; data were measured at
half-hourly intervals with EM50 data loggers. Soil water tension was
measured using 15 cm elbow tensiometers (Soil Measurement Systems, Arizona,
USA) installed horizontally at 5 cm below moss surface; an INFIELD 7
portable tensiometer (UMS AG, Munich, Germany) was used to measure
tension three times per week. Actual evapotranspiration (ET) was measured
directly by weighing the monoliths before and after the rainwater additions
throughout the study period.
CO2 exchange was measured three times per week in the environmental
chamber as above, although under constant atmospheric conditions (Table S1).
CO2 fluxes were measured using a clear Plexiglas cylindrical chamber
(30 cm diameter × 40 cm height). A fan was mounted on the inside of the
chamber to mix the air and homogenize the CO2 concentration during
measurements. NEE was measured by fitting the clear chamber to a monolith
and sealing it with petroleum jelly to ensure an air-tight seal. The chamber was
connected to an infrared gas analyzer (IRGA; EGM-4, PP Systems, Massachusetts, USA). All monoliths were measured sequentially.
Following the NEE measurements, ecosystem respiration (ER) was measured with
an opaque aluminum cover placed over each chamber to exclude photosynthetically active
radiation.
Temperatures inside the clear and opaque chambers were monitored during
measurements and never found to increase by more than 1.5 ∘C.
Gross ecosystem productivity (GEP) was calculated by subtracting ER from
NEE.
Data analysis
The data were checked for homogeneity and normality with a Levene's test and
a Shapiro–Wilk test, respectively. Differences in hydrological and CO2
exchange parameters between rainfall frequency treatments and vegetation
communities were assessed with linear mixed effects models: WT, VMC,
soil tension, ET, GEP, ER, and NPP were response variables; rain fall
frequency and water table treatments were fixed effects (with interactions);
and the repeated measurements of the monoliths were random effect.
Relationships between hydrological variables and CO2 exchange
components were examined with linear and nonlinear regression. All
statistical analyses were performed using the STATISTICA 8 software package
(StatSoft Inc.).
Mean (± standard deviation) values of volumetric moisture
content (VMC), water table (WT), soil matric tension (ψ), and
evapotranspiration (ET) for each precipitation, WT, and vegetation community
treatment during the laboratory experiment. Different letters indicate
significant differences (p<0.05) between precipitation treatments
within each water table and vegetation treatment. No letter indicates no
significant difference.
Rainfall
VMC
WT
ψ
ET
VMC
WT
ψ
ET
treatment
(%)
(cm)
(cm)
(mm d-1)
(%)
(cm)
(cm)
(mm d-1)
High water table
Low water table
Moss
HiFreq-lab
74(8)a
-6(1)a
-18(8)ab
1.4(.07)
50(8)a
-20(4)a
-25(5)a
1.5(0.2)
MedFreq-lab
71(6)ab
-7(2)a
-16(7)a
1.3(.07)
47(6)ab
-20(4)a
-28(6)a
1.3(.07)
LowFreq-lab
65(8)b
-11(3)b
-23(9)b
1.5(0.1)
43(7)b
-23(5)b
-32(9)b
1.4(.08)
Moss + shrub
HiFreq-lab
77(8)a
-8(2)a
-14(7)a
1.4(0.2)
50(7)a
-21(4)
-30(12)
1.6(0.2)
MedFreq-lab
72(5)ab
-10(3)ab
-18(10)ab
1.5(.09)
46(8)ab
-22(4)
-28(10)
1.4(0.1)
LowFreq-lab
68(5)b
-12(3)b
-21(10)b
1.6(0.1)
43(8)b
-22(5)
-33(14)
1.5(0.1)
Sedge + moss
HiFreq-lab
67(8)
-13(4)a
-21(10)a
1.8(0.1)a
44(9)a
-31(8)a
-41(11)a
1.9(0.2)
MedFreq-lab
69(10)
-10(3)a
-22(9)a
1.4(0.1)b
48(6)a
-22(5)b
-29(5)a
1.3(0.3)
LowFreq-lab
64(10)
-19(10)b
-33(15)b
2.0(0.1)a
37(10)b
-37(6)c
-71(38)b
1.8(0.2)
Results
Peatland hydrology under changing rainfall frequency
Decreasing precipitation frequency generally resulted in decreased WT depth,
near-surface VMC, and soil tension for all vegetation types, regardless of
the initial WT position (Table 1). Water table fluctuation between rainfall
treatments for the different vegetation monoliths are shown in the Supplement (Fig. S1).
During the high WT period of the experiment, WT depths were significantly
lower in the LowFreq-lab relative to the HiFreq-lab treatment in all
vegetation communities (p<0.001; Table 1). In the second phase of
the experiment when WTs were reset to -15 cm the WT was significantly lower
in the LowFreq-lab relative to the HiFreq-lab treatments in the sedge + moss
(p<0.01) and moss (p<0.05) monoliths, but not
different in the moss + shrub communities.
Exceedance probability of VMC from the (a) moss, (b) moss + shrub, and (c) sedge + moss
peat monoliths subject to Hi-, Med-, and LowFreq-lab precipitation frequency treatments.
Near-surface VMC followed the same pattern of WT fluctuation between
rainfall treatments throughout the experiment for all vegetation monoliths
(Fig. S2). For all vegetation types VMC was >60 % in the
HiFreq-lab monoliths for more than half the experiment (Fig. 1). In
comparison, the 50th quantile of VMC in the LowFreq-lab treatment was
much lower at 49, 51, and 52 % for the moss (Fig. 1a), moss + shrub
(Fig. 1b), and sedge + moss (Fig. 1b), respectively. Overall, VMC was
significantly higher in the HiFreq-lab than the LowFreq-lab treatment for
the moss and moss + shrub monoliths for both portions of the experiment
(p<0.001; Table 1). In the sedge + moss monoliths there were no
significant differences in VMC between treatments under the high WT period
(p=0.298), but during the low WT period VMC in the LowFreq-lab treatment
was significantly lower than either of the more frequent treatments (p<0.005).
Similar trends were found during the field experiment
(Table S3). There was a significant relationship between VMC and WT for all
vegetation communities (R2=0.77-0.93, p<0.0001; Fig. S3),
with proportionally greater decreases in VMC for a given WT drawdown for
low- versus high-frequency treatments for all vegetation types.
Relationship between daily average VMC and days since the
last rainfall/irrigation event for (a) moss, (b) sedge + moss,
and (c) moss + shrub monoliths. Data for each vegetation type are
separated between high WT and low WT portions of the experiment. Data are
combined between precipitation frequency treatments. Linear regressions are
provided. All regressions were significant (p<0.05).
Increasing the duration since the last precipitation event led to
significant decreases in VMC for all vegetation communities (Fig. 2). The
number of consecutive dry days had a greater effect on VMC declines during
the high WT phase of the experiment for both the moss and moss + shrub
monoliths. The VMC rate of decline decreased in the moss monoliths by 50 %
between the two phases of the experiment, from 1.6 % d-1 during
the high WT period (R2=0.40; p<0.001) to 0.8 % d-1
(R2=0.19; p<0.001) during the low WT phase (Fig. 2a). The
decrease in rate of VMC decline was ∼ 31 % between high
and low WT portions of the experiment for the moss + shrub monoliths (Fig. 2c).
In contrast the rate of VMC decline with increasing consecutive dry
days increased slightly in the sedge + moss monoliths as the WT was
lowered, from 1.6 % d-1 during the high phase (R2=0.25; p<0.001) to 1.8 % d-1
during the low WT phase (R2=0.27; p<0.001; Fig. 2b).
Relationship between daily soil water tension (centimetre of
water) and days since the last rainfall/irrigation event for (a) moss, (b) sedge + moss,
and (c) moss + shrub monoliths. Data for
each vegetation type are separated between high WT and low WT portions of
the experiment. Data are combined between precipitation frequency
treatments. Linear regressions are provided. All regressions were
significant (p<0.05).
Average near-surface soil tension increased (became more negative) with
decreasing rain frequency in all vegetation treatments (Table 1). These
tensions rarely reached -100 cm, the critical level for Sphagnum capillary water
supply, except in the sedge + moss monoliths subject to the LowFreq-lab
treatment during the low WT period, where soil tension frequently was
-150 cm. The monoliths experienced linear increases in tension of ≃1.1,
-1.8, and -1.4 cm d-1
of no rainfall under high WT levels for
mosses (R2=0.28; p<0.001), sedge + moss (R2=0.22; p<0.001),
and moss + sedge (R2=0.36; p<0.001), respectively (Fig. 3).
These rates of tension increase generally
remained the same in the low WT portion of the experiment for the moss and
moss + shrub monoliths (Fig. 3a, c). Conversely, soil tension
increased at a rate of -4.3 cm d-1 without rain in the sedge + moss
monoliths during the low WT phase, more than double the rate under high WT
(Fig. 3b).
Sedge + moss evapotranspiration (ET) was significantly higher under
HiFreq-lab and LowFreq-lab than MedFreq-lab treatments during both the high-
and low WT phases of the lab experiment (p<0.05; Table 1). There
were no significant differences in ET in the moss and moss + shrub
monoliths. There was also no clear trend between ET and the number of days since
rainfall; however, ET exceeded 3.5 mm d-1 for up to 2 days after
rainfall in all vegetation communities, and generally remained below 3 mm d-1
for periods up to 14 days without rain.
Results of the mixed effects models on CO2 exchange parameters
from the monolith experiment. Numerator degrees of freedom for rainfall
frequency, water table treatment fixed effects, and their interaction were
2, 1, and 2, respectively, in all models. The denominator degrees of freedom for
all models was 222.
GEP
ER
NPP
F
P
F
P
F
P
Moss
Rainfall treatment
9.02
<0.001
29.14
<0.001
42.37
<0.001
Water table treatment
2.66
0.104
341.19
<0.001
229.53
<0.001
Rainfall × Water table
1.98
0.141
13.86
<0.001
15.75
<0.001
Moss + shrub
Rainfall treatment
7.38
<0.001
5.06
0.007
20.27
<0.001
Water table treatment
1.81
0.180
42.39
<0.001
239.43
<0.001
Rainfall × Water table
1.77
0.173
0.08
0.926
9.13
<0.001
Sedge + moss
Rainfall treatment
5.82
0.003
3.27
0.04
7.83
0.001
Water table treatment
62.78
<0.001
171.78
<0.001
11.69
0.001
Rainfall × Water table
14.24
<0.001
5.10
0.007
6.70
0.001
Comparison of mean GEP, ER, and NEE
between rainfall treatments within each WT treatment for each vegetation
community: (a) moss, (b) moss + shrub, and (c) sedge + moss.
Error bars represent the standard error of the mean. Negative NEE
represents net CO2 uptake. Different letters indicate
significant differences (p<0.05) between treatments within each WT
period.
CO2 exchange dynamics
The rainfall frequency treatments had an effect on CO2 exchange
dynamics for all vegetation communities, with low-frequency rain eliciting a
response for GEP, ER, and NEE (Table 2). The effect of the high versus low
WT treatments was also very strong (Table 2). Figure 4 illustrates this
rainfall treatment's effect on CO2 exchange in greater detail. Gross
ecosystem productivity from the moss monoliths decreased with decreasing
rainfall frequency under low WT, with rates nearly twice as high in the
HiFreq-lab treatment (-0.101 ± 0.010 mg CO2 m2 s-1)
compared to the LowFreq-lab treatment
(-0.056 ± 0.014 mg CO2 m2 s-1; p=0.017; Fig. 4a). While there were no differences in
moss ER between rainfall treatments during high WT, the LowFreq-lab treatment
led to significantly more respiration during the low WT period than either
higher-frequency treatments (p=0.004). Moss NEE significantly decreased
(less CO2 uptake/more CO2 release) with decreasing precipitation
frequency in both WT treatments. Under the high WT period, moss monoliths
were net CO2 sinks, with a NEE 3.5 times lower in the LowFreq-lab (-0.010 ± 0.008 mg CO2 m2 s-1)
relative to the HiFreq-lab
treatment (-0.038 ± 0.006 mg CO2 m2 s-1; p=0.013).
During the low WT treatment, the mosses switched to net CO2 sources,
with the LowFreq-lab treatment emitting between three and four times more
CO2 (0.123 ± 0.016 mg CO2 m2 s-1) than the Med-Freq-lab
(0.040 ± 0.007 mg CO2 m2 s-1) and HiFreq-lab (0.031 ± 0.004 mg CO2 m2 s-1)
treatments (p<0.001; Fig. 4a).
The presence of shrubs with the moss increased GEP in the LowFreq-lab
treatment during the low WT phase, such that there were no differences in
GEP due to rainfall treatment at this time (Fig. 4b). Additionally, during
high WT the MedFreq-lab treatment for the moss + shrub monoliths led to
more CO2 uptake than the LowFreq-lab treatment. There were no
differences in ER between vegetation treatments for the high WT portion of
the experiment for the moss + shrub monoliths, but lower frequency rain
led to greater ER than the HiFreq-lab treatment during the low WT portion.
The pattern of NEE in response to rainfall frequency for moss + shrub
monoliths was similar that observed with moss – under high WT all three
treatments were CO2 sinks, with a shift to sources under low WT (Fig. 4b).
While there were no differences in rates of NEE between frequency
treatments under high WT (p=0.412), under low WT the MedFreq-lab (0.042 ± 0.007 mg CO2 m2 s-1)
and LowFreq-lab (0.07 ± 0.011 mg CO2 m2 s-1)
were nearly 6 and 10 times greater
sources of CO2 than the HiFreq-lab treatment (0.007 ± 0.008 mg CO2 m2 s-1; p=0.019), respectively.
There was a trend of greater CO2 uptake with decreasing rainfall
frequency under high WT in the sedge + moss monoliths, with LowFreq-lab
resulting in greater GEP (-0.155 ± 0.008 mg CO2 m2 s-1)
than HiFreq-lab (-0.098 ± 0.002 mg CO2 m2 s-1; p=0.03; Fig. 4c).
Under low WT, GEP more than doubled in the HiFreq-lab
treatment to -0.214 ± 0.012 mg CO2 m2 s-1, which was
significantly higher than the MedFreq-lab treatment (-0.146 ± 0.007 mg CO2 m2 s-1; p=0.002).
Under high WT conditions the
lower-frequency treatments had higher rates of ER than HiFreq-lab, but this
pattern disappeared with lower WT conditions. Overall, under high WT there
was no effect of rain frequency on the NEE for the sedge + moss monoliths;
however, under low WT there was significantly greater NEE (more carbon
uptake) during frequent rain (-0.045 ± 0.012 mg CO2 m2 s-1; p=0.002)
as compared to MedFreq-lab (0.008 ± 0.007 mg CO2 m2 s-1) and LowFreq-lab
(-0.009 ± 0.012 mg CO2 m2 s-1), both of which were not statistically different from zero
(Fig. 4c).
Hydrologic controls on CO2 exchange
in S. capillifolium-dominated monoliths,
depicting relationships in (a) GEP, (b) ER, and (c) NEE between rainfall
frequency treatments. Relationships in (a) are unimodal with indicated
correlation coefficients and significance. Relationships in (b) are linear
and are indicated with correlation coefficients. All regressions in
(b) were statistically
significant (p<0.001). Relationships for HiFreq-lab and
MedFreq-lab treatments in (c) are second-order polynomial; relationship for
LowFreq-lab is third-order polynomial with indicated correlation
coefficients and significance.
Controls on CO2 exchange between rainfall frequency
treatments
There were significant unimodal relationships between moss GEP and VMC from
the monoliths, and these relationships varied between the rain frequency
treatments (Fig. 5a). The relationship was strongest for the high frequency
rain treatments, with peak rates of GEP occurring at 65 % VMC. The
strength of the relationship decreased with decreasing rain frequency, as
did the VMC at which peak GEP occurred. There were no significant
relationships between GEP and WT. The position of the WT was highly
correlated with moss ER for all of the rain frequency treatments (p<0.001; Fig. 5b).
As rain frequency decreased, WT declines led to
proportionally greater rates of ER, with the rate of increase in LowFreq-lab
ER being 1.5 times higher than the HiFreq-lab treatment (Fig. 5b). Moss NEE
was strongly controlled by near-surface VMC, with significant quadratic
correlations for the HiFreq-lab and MedFreq-lab treatments, and a
third-order polynomial relationship for the LowFreq-lab rainfall treatment
(Fig. 5c). The switch from net CO2 uptake to a net source occurred in
the moss monoliths at VMCs of 60, 62, and 64 % in the HiFreq-lab,
MedFreq-lab, and LowFreq-lab, respectively. VMC was below these levels for
49 % of the study in the HiFreq-lab treatment, 66 % in the
MedFreq-lab, and 73 % in the LowFreq-lab treatment (Fig. 1). Similar
relationships between moss carbon flux and peatland hydrology between rain
frequency treatments were found in the field experiment (Fig. S5); however,
positive NEE was observed at higher VMC, which occurred for much less of the
study period (20 % in the LowFreq treatment).
Hydrologic controls on CO2 exchange
in C. oligisperma -dominated monoliths, depicting
relationships in (a) GEP, (b) ER, and (c) NEE between rainfall frequency
treatments. Relationship in (a) is unimodal with indicated correlation
coefficients and significance. Relationships in (b) and (c) are linear and
are indicated with correlation coefficients. All regressions in (b) and
(c) were statistically significant (p<0.001).
Under the HiFreq-lab treatment VMC had a unimodal control on rates of GEP
from the sedge + moss monoliths, with peak gross production occurring at a
VMC of 49 % (Fig. 6a). Decreasing rain frequency rendered this
relationship insignificant. Ecosystem respiration was influenced by WT
position in the sedge + moss monoliths, with rates increasing between 57
and 71 % faster with WT drawdown in the LowFreq-lab and MedFreq-lab,
respectively, than for the HiFreq-lab conditions (Fig. 6b). There were
moderate but highly significant (p<0.001) linear correlations
between WT and sedge + moss NEE for the lower-frequency treatments, and a
quadratic relationship existed between NEE and WT in the HiFreq-lab
treatment (Fig. 6c). The trends with the moss + shrub monoliths were
similar to the moss-only monoliths (data not shown). There were no
significant relationships between WT or VMC with GEP or NEE from the shrub
communities in the field, although WT did control ER in these communities (p<0.001).
Overall, increasing the number of days since the last rainfall event led to
a strong increase in NEE (CO2 efflux to the atmosphere) for the moss
monoliths (R2=0.31, p<0.01), a moderate increase for the
moss + shrub monliths (R2=0.18, p<0.001), but almost no
increase in the sedge + moss monoliths (R2=0.05, p=0.044;
Fig. 7). This relationship was linear for the vascular plant communities,
and quadratic for the moss-only monoliths. The moss and moss + shrub
monoliths had a net uptake of CO2 when the duration since the last
event was under 3.5 days; however, after this threshold, there was net
emission of CO2. Conversely, sedge + moss monoliths were
largely sinks of CO2 for up to 2 weeks without rainfall, and
were predicted to have a net emission of CO2 after 15 days between
events.
Discussion
Plant community-mediated response of peat hydrology to precipitation
frequency
It is expected that a shift in the precipitation to larger, more infrequent
events will lead to lower WT levels and drier surface conditions (Knapp et
al., 2008; Piao et al., 2009). We found that decreasing the precipitation
frequency while holding total seasonal rain constant resulted in lower
average WT positions, lower VMC, and higher soil tension in all vegetation
communities (Tables 1, S3; Fig. S1, S2). On one hand, the smaller, more frequent
precipitation events were able to buffer against seasonal WT declines and
maintain more moisture in the near-surface peat layer (Fig. S3). On the
other hand, the larger, less frequent events contributed moisture to deeper
layers, which led to the observed increased rates of VMC decline with
concomitant WT decline (Fig. S3). Therefore, in addition to climate change
leading to increased ET and lower WT positions in peatlands (Whittington and
Price, 2006; Munir et al., 2015), the frequency of rainfall events is likely
to contribute to even drier surface conditions than is currently considered.
Decreasing rainfall frequency allowed for continued near-surface soil
moisture decreases (Fig. 2) and tension increases (Fig. 3). Delivering rain
every 3 days, the current average for the studied peatland (Table S2),
prevented these large soil moisture declines in communities dominated by
Sphagnum moss (Table S3; Fig. S2). However, the presence of sedges
increased ET relative to the moss-only communities. Vascular plant abundance
increases peatland ET, even during periods of limited rainfall (Takagi et
al., 1999; Petrone et al., 2004; Admiral and Lafleur, 2007; Wu et al., 2013;
Takashi et al., 2016). During the first portion of the experiment when WTs
were high, the rate of drying was the same between moss and sedge + moss
monoliths; however, during the low WT portion the rate of drying in the
sedge community became 75 % higher than in the moss-only community (Fig. 2a, b).
This led to over a four-fold greater rate of tension increase in the
presence of sedges (Fig. 3a, b). The low-frequency rain treatment in the
sedge + moss community led to near-surface tensions ≤100 cm,
sometimes after only 4 days without rain. This tension threshold is
generally considered the point at which Sphagnum can no longer effectively
photosynthesize (Price, 1997, Thompson and Waddington, 2008).
In our study -100 cm of tension was reached at a VMC of 37 %, similar to
thresholds found in other peatlands (Price and Whitehead, 2004; Cagampan and
Waddington, 2008). The high- and medium-frequency treatments for all
vegetation communities maintained soil moisture above this value for at
least 92 % of the experiment; however, deviations below this threshold
increased considerably in the low-frequency rain treatment (Fig. 1). The
moss and moss + shrub monoliths subject to low-frequency rain experienced
moss-level VMC <37 % for ∼ 10 % of the
experiment, whereas the sedge + moss monoliths were below this threshold
for 38 % of the experiment. Overall, our results suggest that the
interaction between deeper WTs and less-frequent precipitation regimes
expected with climate change will lead to high tensions that may limit
moisture uptake by Sphagnum mosses in peatlands dominated by sedges.
Relationships between NEE and the number of consecutive dry
days since rainfall for each vegetation community. Negative NEE represents
CO2 uptake. Error bars represent the standard
deviation of the mean. The relationships for the moss and moss + shrub communities
are significant at the p<0.001 level ; the relationship for the sedge + moss communities is significant at the p=0.045 level.
Precipitation frequency – peatland hydrology interactive effects on
CO2 exchange
The precipitation regimes we imposed affected the CO2 exchange of the
vegetation communities. Overall, decreasing rainfall frequency led to
decreased NEE (less storage/greater flux to the atmosphere) from the
peatland communities tested, with differences in NEE for four of the six
vegetation–WT combinations we tested (Fig. 4). Nijp et al. (2014) also found
NEE decreased from three Sphagnum species as precipitation frequency decreased.
Field measurements of CO2 fluxes from peatlands have also documented
decreased GEP and increased ER during extended rainless periods, switching
peatlands to net sources of CO2 during these short periods (Alm et al.,
1999; Lund et al., 2012). Seasonal estimates from our study site confirm
increased flux of CO2 to the atmosphere from moss communities during
low-frequency rain; however, the abundant shrub communities we were able to
test in the field reduced losses to the atmosphere under low-frequency rain,
due to higher GEP from the mature shrubs (Radu and Duval, 2018b).
Lower WTs, characteristic of drought periods, can decrease productivity from
Sphagnum-dominated peatlands through reduced VMC creating a moisture stress
(Alm et al., 1999; Chivers et al., 2009; Potvin et al., 2015), as well as
increase aerobic respiration through increased oxygen diffusion and a
greater depth of aerated soil (Carroll and Crill, 1997; Tuittila et al.,
2004); this results in an overall net loss of carbon to the atmosphere. Our study
demonstrates that low WTs exacerbate the effect of low rain frequency through
greater rates of ER, causing the moss and moss + shrub communities to
switch from net sinks of CO2 to net sources (Fig. 4a, b). The
combination of low WT and low-frequency rain led to greater rates of
near-surface drying (Fig. 2), which would stimulate soil respiration through
increased aeration (Silvola et al., 1996). Additionally, low WTs resulted in
the sedge + moss monoliths, subject to the two lower frequency
treatments, switching from sinks to becoming carbon neutral (Fig. 4c). The lack of NEE
response to a low WT from the sedge + moss community subject to frequent
rain was due to increased GEP, presumably because the high frequency rain
maintained moss photosynthesis, while the sedges increased production with
lower WT (Wu et al., 2013; Potvin et al., 2015).
Carbon assimilation by S. capillifolium was sensitive to rainfall frequency-induced changes
in VMC. As rainfall frequency decreased, peak GEP rates were lower and
occurred at lower VMC, while the strength of the relationship between GEP
and VMC weakened (Fig. 5). Repeated drying and wetting of mosses has been
found to result in lower photosynthetic capacity due to damage to moss
cellular integrity, including degradation of chlorophyll and rupture of cell
membranes (Schipperges and Rydin, 1998). The low-frequency rain events
allowed for greater near-surface VMC variability, which may have contributed
to lower moss GEP at any given VMC. Gerdol et al. (1996) found Sphagnum species,
including the dominant species in our study, S. capillifolium, were unable to resume
photosynthesis after an 11-day period without new water. In our study, the
mosses continued to photosynthesize, albeit at lower rates, after 13 days
without rain, as soil water tensions were weak enough (≥100 cm)
to allow water uptake. Mosses subject to a regime of 6 days between
rainfall events had GEP rates 25 % lower than under a regime of 2 days
between events.
The observed increase in the VMC threshold at which the Sphagnum capillifolium monoliths switched
from NEE sinks to sources as precipitation frequency decreased (Fig. 5a) was
related to the concomitant lower WTs (Fig. S3a). Therefore, decreasing rain
frequency not only lowered near-surface soil moisture but also created an
additional stress for the mosses by reducing the availability of capillary
water due to the lower WT. During the high WT portion of the experiment VMC
was found to drop below these thresholds after 10–11 days without rain;
however, the low WT period was characterized by a near-surface VMC almost
always less than 60 % (Fig. 2a). Strack et al. (2009) found a sink–source
threshold for S. rubellum of ∼ 45 % VMC in the near-surface peat, and
Nijp et al. (2014) found that a VMC of 48 % corresponded to a switch from sink
to source for S. balticum. All three species are lawn- or small hummock-forming species
capable of withstanding low water availability, yet they differ greatly in their
response to VMC. The physiological mechanisms driving this range in moisture
content threshold to NEE among Sphagna species were beyond the scope of our study,
but these differences in species-specific responses to soil water content
are very important for parameterizing peatland ecosystem models (le Roux et
al., 2013; Nijp et al., 2017).
In our study, moss-dominated communities became sources of CO2 to the
atmosphere after less than a week without rain (Fig. 7). This switch from
carbon sink to source was primarily driven by decreased VMC between events
limiting GEP (Figs. S2, 5a). The monoliths dominated by vascular plants were
less affected by rainfall-induced decreases in VMC due to their deeper roots
(Silvan et a., 2004), with the sedge communities maintaining uptake or
carbon neutrality for the maximum 14 consecutive dry days of our experiment
(Fig. 7). In our study region, southern Ontario, seven consecutive dry days
occur quite regularly during the growing season, with one or two 14-day dry
periods typical within a growing season. The frequency of extreme
precipitation events is increasing in the region (Cao and Ma, 2006; Soulis
et al., 2016); however, little research has been conducted on the dynamics
of the intervening dry periods. Most climate models project these dry
periods to increase in both length and recurrence during growing seasons
(Orlowsky and Seneviratine, 2012; Sillmann et al., 2013; Walsh et al.,
2014).
Implications for peatland plant functional types and climate
change
Our study demonstrates that plant functional types will be differentially
affected as rainfall becomes less frequent, particularly if accompanied by
lower WTs as a result of climate change. Sphagnum-only communities experienced a
significant decrease in GEP with less frequent rainfall, while GEP in the
communities with vascular plants was generally unaffected. Additionally,
decreased precipitation frequency had stronger negative implications for S. capillifolium in
the presence of sedges, with near-surface tensions of ≥100 cm
after only 3 days without rainfall when the WT was deep (≥15 cm).
Although we did not measure S. capillifolium productivity separate from the sedges in
the same communities, white, desiccated S. capillifolium capitula were
observed in sedge + moss monoliths receiving less frequent rainfall when the water table was
low. In a companion study we found that low frequency rain led to significantly
more seasonal sedge and ericaceous shrub cover and GEP than high frequency
rain (Radu and Duval, 2018b). Ericaceous shrubs in particular seem to thrive,
with rates of GEP under LowFreq exceeding HiFreq rates 75 % of the time,
generally at multiples of 2–7 under high light conditions (Fig. S4).
Our results show that decreased precipitation frequency will decrease net
CO2 uptake in peatland plant communities dominated by Sphagnum, both in the
presence and absence of sedges and juvenile shrubs. With the lower water
table positions expected as a result of increased ET, we showed that these
communities may switch to CO2 sources, and decreased precipitation
frequency may increase net CO2 release. Furthermore, increased shrub
dominance (Hedwall et al., 2017) is likely to shift the CO2 balance
towards increased CO2 efflux to the atmosphere. Therefore, we
found that decreasing precipitation frequency is likely to lead to a
positive feedback for climate change due to increased net CO2 release
to the atmosphere caused by drier surface conditions and a possible shift to
sedge- and shrub-dominated communities.
Over the longer term, decreased precipitation frequency is likely to affect
the net carbon sink function of temperate peatlands. The reduced NEE of the
Sphagnum-dominated monoliths with decreasing precipitation frequency (Fig. 4)
will progressively limit the storage of carbon in the peat on the annual and
decadal scales. We have recently shown that decreased growing season
precipitation frequency increases sedge and shrub growth (Radu and Duval,
2018b). The incorporation of vascular plant litter into the peat matrix
should increase rates of respiration over and above the short-term results
of the present study due to the greater rates of decomposition relative to
Sphagnum peat (Leifeld et al., 2012; Dieleman et al., 2015; Duval and Radu, 2018).
Predicted lower WT position as a consequence of global warming will result
in lower long-term rates of carbon accumulation in peatlands (Strack et al.,
2006; Peichl et al., 2014; Carlson et al., 2015). Therefore, the combined
effects of lowered WTs, increased vascular plant cover, and decreased VMC due
to decreased precipitation frequency (Fig. 2) are likely to reduce long-term
rates of carbon storage, potentially switching temperate peatlands to carbon
sources.