The effects of temperature and nitrogen and sulfur additions on carbon accumulation in a nutrient-poor boreal mire: Decadal effects assessed using 210Pb peat chronologies

Authors


Abstract

Boreal peatlands are a major long-term reservoir of atmospheric carbon (C) and play an important role in the global C cycle. It is unclear how C accumulation in peatlands responds to changing temperatures and nutrients (specifically, nitrogen and sulfur). In this study, we assessed how the C input rate and C accumulation rate in decadal old peat layers respond to increased air temperatures (+3.6°C) during the growing season and the annual additions of nitrogen (N) and sulfur (S) (30 and 20 kg ha−1 yr−1, respectively) over 12 years of field treatments in a boreal mire. An empirical mass balance model was applied to 210Pb-dated peat cores to evaluate changes in C inputs, C mass loss, and net C accumulation rates in response to the treatments. We found that (i) none of the treatments generated a significant effect on peat mass loss decay rates, (ii) C input rates were positively affected by N additions and negatively affected by S additions, (iii) the C accumulation rate in the uppermost (10 to 12 cm) peat was increased by N additions and decreased by S additions, and (iv) only air temperature significantly affected the main effects induced by N and S additions. Based on our findings, we argue that C accumulation rates in surface peat layers of nutrient-poor boreal mires can increase despite the predicted rise in air temperatures as long as N loads increase and acid atmospheric S remains low.

1 Introduction

High-latitude peatlands represent a major long-term sink of atmospheric carbon (C). Their development since the Holocene has resulted in 270 to 550 Pg of stored organic C, which corresponds to around 30% of the total C stored in terrestrial ecosystems [Gorham, 1991; Yu, 2012]. Due to their importance as a long-term C reservoir, understanding how C accumulates and responds to environmental factors in peatlands is crucial for predicting changes in the global C cycle. However, the impact from altered climate and nutrient conditions on C accumulation rates remains highly debated.

The capacity of peatlands to sequester or release atmospheric C resides in the disequilibrium between carbon sequestration via photosynthesis and its loss through land-atmosphere efflux and discharge C transport [Roulet et al., 2007; Nilsson et al., 2008]. Boreal peatlands generally accumulate C, which indicates that the net primary production is typically greater than combined losses [Clymo, 1984]. However, large-scale changes in climate (e.g., temperature and humidity) and environmental constraints (e.g., drainage and anthropogenically derived nitrogen and sulfur deposition) could alter the current imbalance, variations that could change the role of peatlands as net carbon sinks [Bubier et al., 1999, 2007; Malmer and Wallén, 2004; Bragazza et al., 2006; Larmola et al., 2013; Loisel and Yu, 2013].

A warmer and drier climate might result in increased heterotrophic respiration rates from mires because higher temperatures and lower water tables stimulate microbial processes [Billings et al., 1982; Moore and Knowles, 1989; Malmer and Wallén, 2004]. In addition to effects generated by climate changes, anthropogenic emissions of nitrogen (N) and sulfur (S) associated with industrial processes and combustion might affect C accumulation in mires [Granberg et al., 2001]. The global increase in atmospheric N deposition [Galloway et al., 1995] may involve an increase in N availability in mire ecosystems, which have been argued to accelerate microbial processes and respiration and, consequently, increase the amount of C released into the atmosphere [Vitousek et al., 1997; Bragazza et al., 2006; Galloway et al., 2008]. The deposition of sulfate (SO42−) may also affect the C accumulation in mires by hampering plant production [Fergusson and Lee, 1979] and suppressing methane (CH4) emission [Granberg et al., 2001; Gauci et al., 2004]. Warmer temperatures and increased N deposition have been shown to result in a progressive replacement of Sphagnum mosses by vascular plants [Gunnarsson and Rydin, 2000; Berendse et al., 2001; Turunen et al., 2004; Wiedermann et al., 2007]. The effect of this change in species composition may result in a net reduction in the C accumulation capacity of mires due to the faster decomposition of the vascular plant litter in comparison to Sphagnum mosses [Tolonen and Turunen, 1996; Gunnarsson and Rydin, 2000; Berendse et al., 2001]. However, although N additions may increase easily degradable C increasing the initial decomposition, the increment in primary production associated with higher availability of nutrients and increased vascular plant growth may more than offset the loss of C to the atmosphere [Rochefort et al., 1990; Vitt et al., 2003; Turunen et al., 2004].

The net C balance in mires and its response to changing environmental conditions has been usually assessed using short-term field manipulation experiments (typically < 5–9 years) [Shaver et al., 1998; Bubier et al., 2007] and laboratory incubations [Billings et al., 1982; Moore and Knowles, 1989; Bragazza et al., 2006]. Concerns arise, however, when attempting to extrapolate these results to predict longer-term effects. An alternative to these approaches is C accumulation rates estimated from dated peat profiles. The natural radionuclide 210Pb (T1/2 = 22.3 years) can be used for this purpose by establishing reliable chronologies of peat deposits accumulated over the past 100–150 years [Appleby et al., 1997].

In this study, we used peat cores collected after 12 years of field manipulation in a boreal minerogenic mire located in northern Sweden to assess how increasing temperature and N and S deposition affected the decadal C accumulation rates in the upper 40 cm of the peat. For each plot, C input and mass loss rates were obtained using 210Pb dating in order to assess the potential influence of the different treatments on decadal C fluxes.

2 Site Description

The study was conducted in a Sphagnum-dominated boreal mire, Degerö Stormyr (64°11'N, 19°33'E, altitude 270 m above sea level, area 6.5 km2). This mixed mire system is located in the Kulbäcksliden Research Park in the Vindeln Experimental Forests, approximately 70 km from the Gulf of Bothnia in the province of Västerbotten, Sweden. The total catchment area including Degerö Stormyr is ~10 km2 [Malmström, 1923]. The mire consists of a complex system of interconnected oligotrophic mires with an average depth of 3–4 m and a maximum depth of 7.8 m. A detailed description of the site can be found elsewhere [Granberg et al., 2001]. The climate of the area is cold, temperate, and humid, with a mean annual temperature of 1.2°C, a mean July temperature of 14.7°C, a mean January temperature of −12.4°C, and a mean annual precipitation of 523 mm (climate normal 1961–1990) [Alexandersson et al., 1991].

The experimental site was selected to be as uniform as possible in terms of microtopography, vegetation, and water chemistry. It is located in the central part of the mire, in a poor fen lawn (pH 4.5) [Eurola et al., 1984], with a peat depth about 5 m and an area of 2 km2. The ground vegetation layer consists of Sphagnum balticum Russ., C. Jens., Sphagnum lindbergii Schimp., Sphagnum majus (Russ.) C. Jens., and Sphagnum papillosum Lindb. The dominant vascular plants are Eriophorum vaginatum L., Vaccinium oxycoccos L., and Andromeda polifolia L., with Scheuchzeria palustris L.and Carex limosa L. occurring more sparsely.

3 Material and Methods

3.1 Experimental Design and Field Manipulations

The experimental setup is a full factorial design with two levels each for temperature (T) (T or no T), nitrogen (N) (ambient (2 kg) or 30 kg N ha−1 yr−1), and sulfur (S) (ambient (3 kg) or 20 kg S ha−1 yr−1) (see Granberg et al. [2001] for details). The high levels of N and S represent the deposition levels in southwestern Sweden at the start of the experiment. Each experimental combination was duplicated, including four midpoints with N and S additions one half of those used in the N and S treatments (reported as n and s, respectively), giving a total of 20 experimental plots. Since 1994, the treatments have been manipulated at a plot scale (sized 2 × 2 m and separated by 1 m buffer zones). To increase air and soil temperature during snow-free periods, the plots were covered with perforated plastic film supported by transparent polycarbonate frames. This setup increased the mean daily air temperature at 0.3 m above the vegetation surface by 3.6°C and also involved changes in the photosynthetically active radiation (PAR) and humidity: the PAR was reduced up to 20–25% with no change in the radiation spectrum (325–700 nm), and the difference in relative humidity was < 10% in > 90% of the measurements collected during one summer, with the largest deviations occurring during the night [Granberg et al., 2001; Eriksson et al., 2010a]. Nitrogen was added as ammonium nitrate (NH4NO3) and sulfur as sodium sulfate (Na2SO4) over the summer season. One third of each of these treatments was applied directly after snowmelt and the rest in one-sixth doses every month from June to September. The chemicals were dissolved in 10 L of surface mire water and distributed evenly over the plots using a water can. The control plots—the nontreated plots—were watered with the same amounts of surface mire water.

3.2 Peat Collection and Analyses

One peat core (10 × 10 cm and 45 cm deep) per plot was collected using a long sharp knife. To avoid compaction, a PVC board (10 × 50 cm) was inserted adjacent to the peat core during extraction. Directly after extraction, the peat profiles were cut into ten 4 cm sections, including the topmost living plants. The lengths of the sides of each section were carefully measured, and the sections were placed in separate plastic bags, which were taken to the laboratory (the same day) in cooling bags, and then stored in a refrigerator until further preparation. To determine their dry mass and bulk density, the peat sections were dried at 105°C for 48 h, cooled in a desiccator, and then weighed.

Carbon concentrations of each layer were analyzed in one core from each of the plots using a NCS 2500 elemental analyzer (CE Instruments, Italy). Lead-210 was analyzed for a limited subsample of all plots (n = 11). Lead-210 activities were determined by measuring the activities of its granddaughter 210Po. The elapsed time between core sampling (July 2007) and 210Po analyses (October 2011) assured secular equilibrium with its parent nuclide. 210Po analyses were carried out by the complete dissolution of the aliquot samples by microwave digestion and its deposition on silver discs [Sanchez-Cabeza et al., 1998] using 209Po as a tracer for yield determination. Po sources were measured using Ortec ULTRA-AS Ion-Implanted-Silicon Charged-Particle Detectors (Model U-020-450-AS).

3.3 Dating Peat Cores

The atmospheric or unsupported fraction of 210Pb required to establish the age-depth relations [Appleby and Oldfield, 1978] was obtained by subtracting the 210Pb activity from the deepest peat samples, considered as a supported background value, from the total 210Pb activity. Age-depth relations for the peat profiles were established by applying the Constant Rate of Supply (CRS) model [Appleby and Oldfield, 1978]. In light of concerns about possible Pb mobility in peat [Urban et al., 1990; Biester et al., 2007] and inaccuracies in 210Pb chronologies due to sampling and analytical limitations [Farmer et al., 2006; Olid et al., 2008; MacKenzie et al., 2011], 210Pb ages were validated against a reference surface of 1994 represented by cables originally installed on the peat surface at the beginning of the experiment and now found deeper in the mire due to peat overgrowth. Mean value from three measurements of the position of the cables was used to estimate the mean vertical growth rate. Once validated, the 210Pb-derived ages were used for the plots subjected to the same treatment without any direct 210Pb measurements.

3.4 Modeling Peat Growth

One of the methods to evaluate ecosystem C balance in peat is based on estimating short- and long-term C accumulation rates. A basic assumption underlying this methodology is that the amount of C in a given year (C(t), g C m−2 yr−1) is determined by the balance between net annual inputs (I, g C m−2 yr−1) and losses. Understanding the mechanisms behind net changes in C input and losses requires knowledge about the balance between peat mass production and decay processes. For this purpose, different conceptual models for describing peat growth have been developed [Yu et al., 2001].

The most general model for peat development is expressed as follows:

display math(1)

where m represents the cumulative dry peat mass (g cm−2), p is the mass incorporation rate (g m−2 yr−1) at the surface, and k is the mass loss rate (per year). Different assumptions have been suggested when solving for p and k [Wieder and Lang, 1982; Clymo, 1984; Clymo et al., 1998].

One of the simplest solutions to equation (1) was given by Clymo [1984]. Clymo divided the peat profile into two functional compartments (i.e., the acrotelm lying above the minimum water table depth and the underlying, permanently water-saturated catotelm). In this solution both the peat mass incorporation rate (p) and mass loss rate (k) in each compartment are assumed to be constant, and the cumulative organic mass is expressed as follows:

display math(2)

There are several other more complex models that could be applied to describe peat accumulation and decomposition in mires that account for the nonhomogenous character of the peat, plant root effect, or the ecosystem succession [e.g., Wieder and Lang, 1982; Clymo et al., 1998; Frolking et al., 2001; Yu et al., 2003].

This study focused on the effect of the different treatments applied on the C accumulation rates in the surface peat at Degerö Stormyr during the past 12 years. With this aim, we fitted the cumulative peat mass versus 210Pb-derived ages to the mass accumulation model described by equation (2). For the sake of comparison, other models were also applied [Wieder and Lang, 1982; Clymo et al., 1998; Frolking et al., 2001; Yu et al., 2003]. The applied models provided estimates on mass incorporation rate (p, g m−2 yr−1) and mass loss rate (k, per year) in the upper ~ 40 cm of peat profile. Using the C content in the uppermost peat layer, we estimated the C input (I, g C m−2 yr−1) at the surface for each profile. Next, we estimated C accumulation in the upper peat centimeters by subtracting the C losses from this input (I − kC, g C m−2 yr−1). The cumulative C inventory (C, g C m−2) used for this estimation corresponds to the mean C inventory accumulated over the past 12 years of treatments. From these calculations, we were able to determine the C accumulation rate from the treated compartment buried at depth (i.e., decadal or recent C incorporation rate).

Because the obtained values are average values for the combined treatment periods, given the assumption of near-steady state conditions, they cannot be linked to specific changes in plant cover during the treatments. The steady state assumption seems reasonable considering that plant community observed at Degerö after 10–12 years of treatments is identical to those observed in mires exposed to anthropogenic deposition for more than 30 years [Wiedermann et al., 2007, 2009; Eriksson et al., 2010a, 2010b]. Furthermore, the absolute numbers derived using the applied models should be interpreted with caution because the applied models assume an input of organic matter only at the peat surface; hence, these models do not consider root inputs as an important source of carbon below the peat surface [Saarinen, 1996; Moore et al., 2002]. Considering these constraints of the model, we initially interpreted the data only in a relative manner to identify altered processes in response to the different treatments.

3.5 Statistical Analyses

Using three-way analysis of variance (ANOVA), we analyzed the effects of temperature (two levels: ambient and + 3.6°C), nitrogen (three levels: 2, 15, and 30 kg ha−1 yr−1), and sulfur (three levels: 3, 10, and 20 kg ha−1 yr−1) and their interactions on peat properties (groundwater table, 210Pb inventories, and peat growth rates) and model parameters (mass loss rate, C input, and C accumulation rates). When needed, data were Log10(x + 1)-transformed to fulfill assumptions of normality and homogeneity of variance. When the interaction among effects was significant, single effects were analyzed; otherwise, Tukey post hoc was used to identify treatment effects. Midpoints (i.e., half addition of the amount used for the N and S treatments) were only analyzed at an ambient temperature. Statistical differences were considered significant for P values lower than 0.05.

4 Results

4.1 Vertical Distribution of Peat Density and C Content

The peat density and C content in the cores are shown in Figure S1 in the supporting information. In short, N additions (N, N × S, N × T, and N × S × T plots) seemed to increase peat density in the upper 12 cm in comparison to the other treatments. Similarly, N additions appeared to have also increased the C content especially in the uppermost 26 cm, with values up to ~ 48–50% for both N and combined N × S treatment plots. For T and S treatments, the C content was fairly constant (44–45%), with values similar to the values for control plots. Similar trends in C content were observed for the combined S × T plots.

4.2 Lead-210 Dating: Vertical Peat Growth

Lead-210 activities decreased exponentially at a depth of 307 ± 64 Bq kg−1 (mean ± standard deviation) in the uppermost peat (2 cm) to 52 ± 23 Bq kg−1 at the bottom of the profiles (38 cm) (Figure 1). This value was taken as supported 210Pb and was, therefore, subtracted from the total 210Pb activities to obtain the unsupported or excess 210Pb fraction. No statistical differences (P > 0.05) in the unsupported 210Pb inventories between the cores were observed (Table 1).

Figure 1.

Vertical distribution of 210Pb activities for the analyzed profiles.

Table 1. Water Table Depth, Unsupported or Excess 210Pb (210Pbxs) Inventories, and Vertical Peat Growth in the Peat Compartment (i.e., Post-Treatment and Pre-Treatment Peat)a
    Vertical Growth (cm yr−1)
TreatmentCoreWater Table Depth (cm)210Pbxs Inventory (Bq m−2)(Post-Treatment)(Pre-Treatment)
  1. a

    Different letters (b and c) in the same column indicate a significant difference (P < 0.05). Mean values are indicated in italics. Standard deviations are indicated with ±. Statistical differences (P < 0.05) identified using a three-way ANOVA are indicated by small letters above mean values.

Control11b12.2b2131 ± 53b0.9 ± 0.2b0.4 ± 0.1
 19b16.2b2190 ± 64b1.1 ± 0.2b0.4 ± 0.1
  b14.2 ± 2.8b2161 ± 42b1.0 ± 0.1b0.4 ± 0.1
N12c5.6b1547 ± 57c0.50 ± 0.06b0.4 ± 0.1
 18c6.6b2020 ± 60c0.54 ± 0.07b0.4 ± 0.1
  c6.1 ± 0.7b1784 ± 334c0.52 ± 0.03b0.4 ± 0.1
S3b12.5b1993 ± 54b0.8 ± 0.1b0.33 ± 0.08
T4b12.2b2223 ± 67b1.1 ± 0.4b0.4 ± 0.1
 16b12.1b1407 ± 103b0.8 ± 0.2b0.4 ± 0.2
  b12.15 ± 0.07b1815 ± 577b1.0 ± 0.2b0.4 ± 0.2
N × S8b5.8b2699 ± 58b0.53 ± 0.03b0.4 ± 0.1
N × T17b11.9b2156 ± 66b0.7 ± 0.1b0.4 ± 0.1
S × T10b12.5b1724 ± 58b0.9 ± 0.3b0.37 ± 0.09
N × S × T2b10.0b2505 ± 85b0.33 ± 0.03b0.34 ± 0.08

The mean vertical peat growth rates were estimated using the CRS-dating model (Table 1). For each 210Pb-dated core, we estimated the mean growth rate corresponding to the upper ~ 8 to 12 cm (i.e., peat developed during the treatment period, i.e., the last 12 years) and the corresponding one below (i.e., nontreated peat). The results showed a mean vertical growth of 1.0 ± 0.1 cm yr−1 for control plots. Similar (P > 0.05) vertical growth was obtained for S, T, and S × T treatments, with 0.8 ± 0.1, 1.0 ± 0.2, and 0.9 ± 0.3 cm yr−1, respectively. The 210Pb-modeled vertical growth rates (average 0.8 cm yr−1) appear highly valid considering that there was no statistical difference (P > 0.2) with those calculated using the levels of the cables placed on the peat surface in 1995 (average 0.9 cm yr−1). Additions of N, however, had a significant effect (P = 0.026) on the vertical peat growth rate with a lower value 0.52 ± 0.03 cm yr−1. In the older pretreatment peat layers (> 12 yr), there was no significant (P > 0.05) difference in peat growth rates between the treatments and all cores had similar average vertical growth rate values around 0.4 cm yr−1.

4.3 Peat Development and C Accumulation Rates in Surface Peat

To calculate C accumulation rates in the upper peat, we modeled cumulative mass versus age for each treatment using equation (2) for the 210Pb-dated peat cores (n = 11) (Figure 2). All points were used to obtain best fit values for C input (p) and mass loss rates (k) from the accumulation model. For all profiles, a good fit was obtained when equation (2) was used to describe the vertical accumulation of mass. A good fit was also obtained when more sophisticated models were applied [Wieder and Lang, 1982; Clymo et al., 1998]. However, the values for p and k derived from these models were similar to those obtained using equation (2). These results together with the better fit obtained from the single exponential model suggested limited use of the more complex models, so we chose the single exponential model to assess C accumulation in all the cores. The estimated mass input (p) and mass loss rates (k) for the 210Pb-dated cores were used for the plots subjected to the same treatments without any direct 210Pb measurements, estimating the C accumulation rates at the upper peat layers (Table 2).

Figure 2.

Cumulative mass versus 210Pb ages in the analyzed profiles. Number of replicates (n) per plot is indicated in each graph. Curves are obtained by fitting the data to the mass accumulation model given by equation (2).

Table 2. Modeled C Input (I, g C m−2 yr−1) and Constant Rates (k, yr−1) From Decadal Modela
  NS Decay Constant (k)Cinput (I)Caccum (I-kC)
TreatmentCore(kg ha−1 yr−1)(kg ha−1 yr−1)T(yr−1)(g C m−2 yr−1)(g C m−2 yr−1)
  1. a

    Different letters in the same column indicate a significant difference (P < 0.05). Reported uncertainties are the 95% confidence interval for the curve-fit parameters. These parameters were used to estimate decadal C accumulation rates (I − kC) for the peat accumulated over the past 100–150 years (temporal scale covered by 210Pb). Standard deviation values are indicated with ±, and small letters above mean values indicate statistical significant differences between treatments (P < 0.05).

Control1123Lowb0.0158 ± 0.0013b113 ± 11b94 ± 11
 1923Low b118 ± 11b98 ± 11
N12303Lowb0.0077 ± 0.0011c108 ± 19c99 ± 19
 18303Low c111 ± 20c104 ± 20
S3220Lowb0.0103 ± 0.0015c104 ± 18b92 ± 18
 7220Low c103 ± 18b88 ± 18
T423Highb0.0119 ± 0.0010b102 ± 11b89 ± 11
 1623High b104 ± 11b92 ± 11
N × S83020Lowb0.0110 ± 0.0008c124 ± 10b108 ± 10
 143020Low c136 ± 11b123 ± 11
n × s11510Lowb0.0110 ± 0.0008c123 ± 10b110 ± 10
 61510Low c128 ± 10b120 ± 10
 151510Low c130 ± 10b116 ± 11
 201510Low c125 ± 10b115 ± 10
N × T9303Highb0.0167 ± 0.0014b139 ± 13b111 ± 13
 17303High b148 ± 14b122 ± 14
S × T5220Highb0.0145 ± 0.0004c120 ± 4c95 ± 4
 10220High c116 ± 4c98 ± 4
N × S × T23020Highb0.0098 ± 0.0004c97 ± 5c87 ± 5
 133020High c101 ± 5c91 ± 5

There was no significant effect on mass loss constant (k) in response to the treatments (P > 0.765) (Table 3). Calculated C input rates (I = p × C% of the upper 4 cm of each profile, g C m−2 yr−1) and C incorporation rates in the upper peat layers (I − kC, where kC means the C losses during the past 8–12 years of treatments, g C m−2 yr−1) are shown in Table 2. Significant main treatment effects on C input rates were observed for the N and S treatments (P < 0.001 and P = 0.029, respectively). Significant effects of the interaction terms of these two treatments (N × S) and treatments involving temperature (S × T and N × S × T) were also observed (Table 3). A significant effect on C accumulation was observed only for high nitrogen deposition plots (P < 0.001) and the combined treatments S × T (P = 0.014) and N × S × T (P < 0.001) (Table 3). Plots subjected to an increase in N load had higher C accumulation rates (N = 101 ± 4, N × S = 116 ± 11, n × s = 115 ± 4, N × T = 117 ± 8 g C m−2 yr−1) in comparison to the C accumulation observed in control plots (control = 96 ± 2 g C m−2 yr−1). Only the combination N × S × T significantly reduced the amount of C sequestered (89 ± 3 g C m−2 yr−1).

Table 3. Three-Way ANOVA Results (F Ratios, and Significance Levels) for the Effects Induced by Additions of Nitrogen (N) and Sulfur (S) and Modified Air Temperatures (T)a
 kCinputlog(Caccum + 1)
TreatmentFPFPFP
  1. a

    Dependent variables are mass loss rates (k), C inputs (Cinput), and C accumulation rates (Caccum) in each treatment.

N0.0150.90424.467<0.00127.809<0.001
S0.0120.9166.2820.0292.2260.164
T0.0190.8930.3970.5411.4000.262
N × S<0.0010.99010.9960.0072.0020.185
N × T0.0160.9020.0340.8572.0020.185
S × T0.0010.97020.9150.0018.6000.014
N × S × T0.0940.765123.994<0.00134.098<0.001

5 Discussion

5.1 Reliability of the 210Pb Chronology

Lead-210 has been widely used to establish the chronology of recent (past 100–150 years) peat accumulation. However, the reliability of this method has been questioned [Urban et al., 1990; Biester et al., 2007; Klaminder et al., 2011]. For this reason, the integrity of the 210Pb record and the reliance on the 210Pb ages should be validated before the outcome of the peat mass and C accumulation modeling is interpreted.

The reliability of the 210Pb chronology in this particular study is supported by three independent results. First, the inventories of atmospheric 210Pb in all the cores collected in Degerö Stormyr (2046 ± 410 Bq m−2) are consistent with those obtained from soils adjacent (less than 60 km away) to the mire (average 2500 Bq m−2) [Klaminder et al., 2006], finding that contradicts the criticism that inaccuracies in peat ages are due to incomplete 210Pb inventories [Farmer et al., 2006; Olid et al., 2008; MacKenzie et al., 2011]. Second, the modeled growth rates appear highly valid considering that there was no statistical difference (P > 0.2) between the 210Pb-modeled growth rates (average 0.8 cm yr−1) and those calculated using the levels of the cables placed on the peat surface in 1995 (average 0.9 cm yr−1). Third, modeled mass incorporation rates derived from Clymo's [1984] model for the control plots (261 ± 24 g m−2 yr−1, Figure 2) is directly comparable to litter-addition rates reported in other studies (mean of 265 ± 37 g m−2 yr−1 [Yu et al., 2001]). In addition, the derived mean decadal mass loss rate (0.016 ± 0.001 yr−1) agreed with previous reported values for boreal forest and wetlands (range from 0.0001 to 0.045 yr−1) [Trumbore and Harden, 1997; Moore et al., 2002; O'Donnell et al., 2011].

5.2 Effects on Mass Loss, C Input, and Surface C Incorporation Rates

Our results suggest that none of the treatments have significantly affected the mass loss of peat, but the addition of N and S substantially affected the input rate and subsequently also the C incorporation rate (Table 3). Temperature affects only the C input and C incorporation by modifying the effects from N and S (i.e., the S × T and N × S × T interaction terms). As the results suggest, N has a generally positive effect on the C input and the C incorporation rate, whereas S has a negative effect.

Our results support previous findings indicating that the N inputs increase the mire's net uptake of C at a decadal time scale [Humphreys et al., 2006; Lund et al., 2010]. However, these results also contradict several previous studies that have argued that an increase in N deposition likely reduces C accumulation rates in mires as N deposition may change plant composition from Sphagnum communities, which produce litter that slowly decomposes, to vascular plants, which produce more rapidly decomposing litter [e.g., Tolonen and Turunen, 1996; Gunnarsson and Rydin, 2000; Berendse et al., 2001; Turunen et al., 2004]. In the N addition plots, a dramatic reduction in Sphagnum communities and an expansion of vascular plants has been observed since the onset of the experiment [Wiedermann et al., 2007, 2009; Eriksson et al., 2010a], but our findings indicate that this replacement has increased the mire's capacity to incorporate C into its surface peat layers rather than reducing the C incorporation rate. The most apparent mechanism to explain these results is the higher gross primary production that is expected to accompany N fertilization of nutrient-poor boreal bogs [Aerts et al., 1995; Vitt et al., 2003; Lund et al., 2010]. According to our modeling efforts, the increase in C input (I) in the N plots is more pronounced than any eventual effect on the mass loss rate occurring as a result of changing litter quality; therefore, our results serve as evidence for a C accumulation rate in the upper part of nutrient-poor boreal mire that is mainly regulated by variable plant input rates rather than by variable litter decomposition rates [Charman et al., 2013; Loisel and Yu, 2013]. In the S treatments, however, the observed negative effect on C input rates appears to be too small to substantially affect C accumulation rates in decadal old peat layers (Table 3). Nevertheless, the negative effect on C input rate seems logical as plants are negatively affected by acid sulfur additions [Longton, 1988; Mäkipää, 1995]. This finding is reinforced by the observed reduction in the sedge cover of about 10% in response to sulfur deposition in this mire [Eriksson et al., 2010a].

An increase in annual air temperatures may accelerate peat decomposition and reduce the C accumulation rates in mires [e.g., Billings et al., 1982, 1983; Dorrepaal et al., 2009]. However, temperature had no main effect in the treatments and only influenced the effects on the C input and C incorporation rates induced by N and S additions, so 12 years of increased annual air temperatures by 3.6°C did not generate a substantial reduction in C accumulation in decadal old peat layers (Table 3). Our results are consistent with those observed by Wieder et al. [1994]: C accumulation rates did not decrease in response to an annual air temperature increase of 4.5°C. In addition, bryophyte production has been shown to respond weakly to increased temperatures [Weltzin et al., 2001]. Therefore, the weak response in the decadal C accumulation in the T treatments was expected and indicates that environmental constraints other than air temperature are likely to control peat production and mass losses in the studied mire. Given the importance of the water table for decomposition processes, it seems important to state that the experimental setup used in this study allows only measurements of direct air temperature effects on C accumulation rates in decadal old peat layers, so we have no measures of effects from the combined impact from increased air temperature and altered groundwater tables. Therefore, we do not question previous studies that concluded that the C accumulation capacity of mires might be reduced in the near future due to increasing heterotrophic respiration rates induced by the combined influence of increased temperature and lower groundwater tables [Gorham, 1991; Christensen et al., 1999; Dorrepaal et al., 2009]. Furthermore, the combined uncertainties from peat heterogeneity and carbon modeling, generating uncertainties of several g C m−2 yr−1 as indicated by the within treatment variability (Table 2), make it difficult to detect small effects by the treatments of this magnitude. Therefore, we cannot exclude the possibility that there is a small main effect induced by the T treatment on the measured parameters (C input, mass loss, and the C incorporation rate), which occurs at a magnitude of a few g C m−2 yr−1.

Importantly, the significant effects seen for the interaction terms between N, S, and T means that predictions of future changes in C input and decadal incorporation rate in nutrient-poor boreal mires need to consider future N and S deposition rates as well as future air temperatures. For example, the positive effect on C input and C incorporation rates induced by N additions is most pronounced when high N additions are accompanied by increased temperatures. That is, the increase in C incorporation rates is significantly higher in the N × T treatment than in the N treatment. Similarly, the effect from increased N additions is expected to be nonsignificant if these additions simultaneously occur with high S loads. Nevertheless, N addition seems to be the single most important factor for the C incorporation rate as indicated by a significant positive effect by this factor alone even when excluding the variable T from the analysis (P = 0.016). Interestingly, significant effects on decadal C incorporation rates are already observed between ambient N additions (2 kg ha−1 yr−1) and midpoint N additions (15 kg ha−1 yr−1). These results suggest that a modest rise in N additions is enough for a significant positive effect on C incorporation rates in upper peat layers.

5.3 Implications

Nitrogen deposition rates are predicted to increase over the next several decades, reaching depositional fluxes up to 2 times higher than current rates [Galloway et al., 2004]. It seems likely that such an increase in N deposition has the capacity to increase the carbon sequestering capacity significantly of nutrient-poor boreal mires considering that such a twofold increase has generated increased C accumulation in decadal old peat layers in our manipulated site. However, if cooccurring with high loads of S and an increase in annual air temperatures, this effect might be reduced, an outcome that is suggested by the negative effect on C accumulation rates in the combined N × S × T (Tables 2 and 3). However, a future scenario with higher S load seems very unlikely for boreal regions considering that acid S deposition has decreased during the last three decades in Europe and North America. Instead, our findings suggest that current and future C accumulations in boreal peat layers induced by increased N levels are likely to be greater than during the 1970s and 1980s, when acid S deposition rates were generally higher.

Recent predictions suggest that at the end of the 21st century large parts of the boreal region are expected to experience increased annual air temperatures comparable to the increased temperatures generated in our treatments (+3.6°C) [Intergovernmental Panel on Climate Change, 2013]. Our findings suggest that the effects from this climatic change on the C cycle in nutrient-poor boreal mires cannot be fully understood without simultaneously considering future N deposition scenarios.

Acknowledgments

This study was financed by Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (grant 21.4/2003-0876) to M.N. and the Swedish Research Council (project 2009–3282) and by Umeå University via a Young Researcher Award to J.K. The authors wish to thank Patricia Rodríguez for her help with the statistical analyses.

Ancillary