Introduction

Half of the Earth’s land surface is influenced by large mammalian herbivores, livestock or native herbivores (Olff et al. 2002), which makes it important to consider the influence they have on the ecosystem. Removal of above-ground biomass by consumption of herbivores will change the below-ground carbon stock and the reduction of litter that goes into the ground will influence the total ecosystem carbon content (Mulder 1999; Sjögersten et al. 2011; Tanentzap and Coomes 2012; Van der Wal et al. 2007). In the Arctic, herbivory has been shown to have an important impact on the carbon cycle (e.g., Cahoon et al. 2012; Mulder 1999; Olofsson et al. 2004; Sjögersten et al. 2008, 2011; Speed et al. 2010; Van der Wal et al. 2007; Welker et al. 2004). The arctic ecosystems are an important global carbon sink and despite that the northern permafrost region only covers about 16 % of the earth, it stores approximately 50 % of the global below-ground organic carbon (McGuire et al. 2009; Ping et al. 2008; Post et al. 1982; Tarnocai et al. 2009). Arctic wetlands in particular are holding large amounts of carbon, as the decomposition rate of organic matter is slow under cold and anoxic conditions (Tarnocai et al. 2009). Anoxic decomposition and methanogenesis and thereby CH4 production prevails under wet conditions. Artic wetland ecosystems produce approximately 40 % of the natural global emissions of CH4 (Cicerone and Oremland 1988; Mikaloff Fletcher et al. 2004).

Plant-soil-herbivore interactions are complex, and involve both direct and indirect impacts, and may influence a variety of ecosystems processes, such as carbon sequestration, greenhouse gas production and emission, vegetation species composition, soil physical parameters (i.e., soil moisture and soil temperature), decomposition rate and nutrient availability (e.g., Sjögersten et al. 2008; Tanentzap and Coomes 2012). Depending on the ecosystem and grazing pressure, herbivory may either lead to an increase in net primary production (NPP) (Cargill and Jefferies 1984; Olofsson et al. 2001, 2004) or a decrease in NPP (e.g., Bagchi and Ritchie 2010; Beaulieu et al. 1996; Cahoon et al. 2012; Ouellet et al. 1994; Sjögersten et al. 2011; Susiluoto et al. 2008; Van der Wal et al. 2007). In nutrient poor high arctic areas a number of studies have shown an increase in NPP as a result of increased herbivory (Olofsson et al. 2004; Van der Wal et al. 2004) as nutrient addition by animal excrement can increase the labile nutrient level (Stark et al. 2002; Van der Wal et al. 2004). Most studies on the impacts of grazing have however reported no effect or a decrease in aboveground NPP (Milchunas and Lauenroth 1993). In some cases grazing can change the carbon balance from being a carbon sink when not influenced by herbivores to becoming a carbon source when grazed (Sjögersten et al. 2011; Welker et al. 2004).

Additionally, studies of grazing in the high arctic have demonstrated that herbivory can result in a shift in vegetation composition—towards being more herb and graminoid dominated (Cahoon et al. 2012; Olofsson et al. 2009; Ouellet et al. 1994; Post and Pedersen 2008; Sjögersten et al. 2008; Stark et al. 2002; Susiluoto et al. 2008; Van der Wal 2006). Grazing has also been shown to change nutrient allocation patterns, as vegetation uses carbon and nutrients reserves for regrowth of new plant shoots instead of building reserves below-ground (Beaulieu et al. 1996; Chapin 1980; Green and Detling 2000; Mulder 1999). Compared to other habitats in the arctic, the mires is highly productive (Arndal et al. 2009) and they are generally more exposed to grazing, as these habitats are preferred by many herbivores. Consequently, they may have a higher carbon loss, than mesic habitats (Sjögersten et al. 2008; Speed et al. 2010).

Despite the potentially large impact of herbivory on the carbon cycle of arctic wetlands and on many of the controlling aspects for CH4 production and emission, only few studies have focused on the effects of herbivory on CH4 fluxes in the arctic and sub-alpine regions (Sjögersten et al. 2011, 2012, respectively). These studies showed no effect of herbivory on CH4 emissions (Sjögersten et al. 2011, 2012). The arctic study was however, focusing on barnacle geese (Sjögersten et al. 2011) and the sub-alpine study were performed in a much drier habitat (Sjögersten et al. 2012), so comparability of the studies is limited.

Many factors are known to influence the CH4 flux; these include soil temperature, water table depth (Torn and Chapin 1993; Waddington et al. 1996), substrate availability and quality (organic acids) (Christensen et al. 2003; Joabsson et al. 1999; Ström et al. 2003), and the presence of certain vascular plant species e.g. Eriophorum scheuchzeri (Ström and Christensen 2007; Ström et al. 2003, 2012). During plant growth, low molecular weight organic acids are released to the rhizosphere, and of these, acetic acid is a substrate of particular importance for methanogenesis. The organic acids are generated from two main sources: (1) from fermentative microbes producing organic acid (OA) (e.g., acetic and formic acid) from plant residues (Charlatchka and Cambier 2000; Gounou et al. 2010; Killham 1994), and (2) from root exudation (Kuzyakov and Domanski 2000; Ström et al. 2003). Despite that organic acids generally accounts for less than 10 % of the dissolved organic matter (DOC), they are essential for the carbon biogeochemistry and nutrient cycle in the soil, as they are bioavailable (Fischer et al. 2007). Hence, to understand the potential effects of herbivory on CH4 emission it is vital to increase our knowledge of the effects grazing have on below-ground substrate availability.

This study focuses on how increased grazing pressure effects the carbon balance, in particular CH4 emissions, in a high arctic mire. We hypothesis that increased grazing will: (1) play an essential role in the carbon cycle and will decrease net ecosystem exchange (NEE) and gross primary production (GPP); (2) decrease substrate availability for CH4 production as fixed carbon will be allocated primarily to above-ground regrowth and to a lesser extent to the root system; (3) lower below-ground C-allocation will lead to decreased CH4 fluxes in plots where increased grazing is simulated; and (4) change the vegetation composition/density. In order to test our hypotheses, an experiment in a high arctic mire, that already is exposed to muskox grazing (Kristensen et al. 2011), was conducted where increased muskox grazing were simulated by clipping the vegetation. Over a 3 year period we measured CO2 fluxes, CH4 fluxes, the OA concentration in pore-water and the ecosystem properties (water table depth, active layer depth and soil temperature). Additionally, we examined the changes in the vegetation composition induced by increased grazing pressure. Though large herbivores, such as muskoxen, may affect the ecosystem both by grazing, trampling and adding nutrients to the ecosystem (e.g., Tanentzap and Coomes 2012), the present study only considers the effect caused by removal of additional vegetation following clipping.

Materials and methods

Site description

The study took place in the arctic valley Zackenberg in NE Greenland (74°30′N 20°30′W). The area is situated in the high arctic, with an annual mean temperature around −9 °C, the warmest month (July) has a mean monthly air temperature (MMAT) of 5.8 °C, and in the coldest month (February) MMAT is −22.4 °C. The mean annual precipitation was 261 mm in the period 1996–2005, with only 10 % falling as rain during summer (Hansen et al. 2008). The area is in a zone with continuous permafrost, and the active layer thickness (i.e. the upper layer of the soil that thaws every summer) varies from 45 to 80 cm depending on the type of area (Christiansen et al. 2008). There are five dominating plant communities classified in the valley: Mire, Grassland, Salix snow-bed, Cassiope heath and Dryas heath (Elberling et al. 2008). The measuring site is located in the freshwater lowland mire Rylekæret. The mires cover approximately 4 % of the valley (Arndal et al. 2009). This ecosystem is normally water-saturated throughout the growing season, years with little snow can, however, lead to a drying during the growing season. pH is relatively high with values around 6.9 ± 0.2 (Ström et al. 2012). The dominating vascular plant species are the three sedges Carex stans, Dupontia psilosantha and Eriophorum scheuchzeri (Christian Bay, personal communication, and our data). Underneath the sedges a dense moss cover is found, e.g., species of Tomenthypnun, Scorpidium, Aulacomnium and Drepanoclaudus (Ström et al. 2012). The peat layer at the measured site is between 18 and 20 cm deep (Falk and Ström unpublished results).

The muskox Ovibos moschatus is the only large herbivore in Northeast Greenland and is present in the Zackenberg area all year-round. During summer, muskoxen predominantly feed in the grasslands and mires mainly eating graminoids (Kristensen et al. 2011), in winter they prefer areas with thin snow-cover, where it is easier for them to access the vegetation (Berg et al. 2008). When muskoxen feed on graminoids, they press the incisors against the pad and pull, leaving the vegetation cut just a couple of centimeters above the surface (Kristensen et al. 2011). Bliss (1986) estimates that muskoxen in general consumes 1–2 % of the sedge dominated meadow cover per year, but may in some areas consume up to 20 % of the available vegetation (Jefferies et al. 1994).

Experimental setup

In 2008, five replicate blocks with three experimental treatments were established in the mire. The replicate blocks were placed closely in a homogeneous part of the mire. The habitat was determined as homogeneous based on a previous study by Ström et al. (2012) that showed very low variation in active layer, soil temperature and water-table depth between 15 plot replicates positioned in the same part of the mire. The three treatments are described below.

  • One plot in each block served as an un-manipulated control, the plot was however still exposed to ambient muskox grazing and trampling.

  • A second plot in each block had the vegetation cut approximately 3–4 cm above the surface twice each summer using a scissor, thus mimicking muskox grazing in the mire (see Kristensen et al. 2011). The vegetation was first cut when it reached a stabled height in the beginning of the growing season, whereas the second cut was made when the grazed vegetation had reached the surrounding vegetation height again. The plots were cut 2 July and 17 July in 2010, 15 July and 1 August in 2011 and 19 July and 13 August in 2012.

  • A third plot in each block had all vascular plants removed in 2008, leaving only the mosses. New shoots were thereafter removed each summer. The vascular plants and their roots were gently pulled from the plots as soon as they became visible. These plots thus mimic the herbivore-induced change into moss-dominated plant communities observed by Sjögersten et al. (2008). No vascular plants (NV) plots represent extreme grazing and were initially established to investigate what happens with the CH4 emission and the OA concentration when vascular plants are absent.

Muskoxen are moving freely in the area and all plots are exposed to both grazing and trampling. By clipping one-third of the plots twice each summer, we increase the grazing pressure substantially.

Before the muskox grazing simulation experiment was initiated, the plots were measured twice in 2009 and twice in 2010 in order to get an indication of the pre-experimental difference between control and clipped plots (later selected through randomization).

Each plot consisted of an aluminum base (39.5 × 39.5 cm) permanently installed (in 2008) 15 cm into the ground. Following the initiation of the simulated grazing in 2010 CH4 and CO2 fluxes and ecosystem parameters were measured on these plots approximately twice per week over the main part of the growing season and in the autumn in 2012. The measurements took place between 10 am and 5 pm, the specific time of the measurement varied each measuring day. No measurements were performed under windy conditions (when wind speed exceeded around 10 m s−1). In 2010 the blocks were measured on 10 occasions between 21 June to 7 August and in 2011 on 13 occasions between 23 June and 5 August. In 2010 and 2011 we observed that the growing season was far from over in the beginning of August (see Figs. 3, 7). Consequently, in 2012 we prolonged our measurement period and the blocks were measured on 23 occasions between 1 July and 18 October.

Flux measurements

The fluxes of CO2 and CH4 were measured using a closed chamber technique (Christensen et al. 2000; Ström and Christensen 2007). For each measurement, a light and a dark measurement was made to establish NEE and the ecosystem respiration (Reco) respectively. Measurements were performed with a transparent Plexiglas chamber equipped with a metal frame with a rubber list, to ensure an airtight seal against the aluminum base. The chamber had a volume of 0.041 m3, and the area of the aluminum base was 0.156 m2. The Plexiglas chamber reduced PAR by <10 % (Christensen et al. 2003). Dark measurements were conducted with the same chamber covered by a non-transparent plastic hood. Immediately before the start of a measurement the chamber was carefully placed on the aluminium base to avoid disturbance and each measurement lasted between 3 and 7 min. Chamber was equipped with a small fan to assist with the circulation and therefore mixing of the chamber air during the measurements. Air pressure inside the chamber was equalized by a small hole in the chamber, which was closed with a rubber stopper as soon as the measurement began.

In 2010 CH4 fluxes were measured by a LGR RMT-200 Fast Methane Analyser (DLT200, Los Gatos Research, USA) and an infrared CO2 gas analyser (PP-systems SBA-4, EGM-4, Hitchin, Hertfordshire, UK), the accuracy of both instruments are 1 %. The air from the chamber was pumped with a rate of 0.4 L min−1 to an analytical box containing the analyzers before it was non-destructively returned to the chamber. In 2011 and 2012, gas concentrations of CO2 and CH4 were simultaneously measured by a portable Fourier transform infrared (FTIR) spectrometer (Gasmet Dx 40-30, Gasmet Technologies Oy). The air from the chamber was pumped at a rate of 3.4 L m−1. Both instruments was set up to record the CH4 and CO2 concentration every second. The FTIR was calibrated with a zero gas every second week. Stable high quality measurements with minimal baseline drift require a cell temperature of the FTIR between 20 and 35 °C. During summer optimal conditions were kept by the internal heating system of the FTIR and during the cold conditions in September and October 2012 the instrument was warmed with heating cables. To validate that the measurements from the two instruments were comparable, a control of the concentration of CH4 measured in ambient air was performed prior to the field season of 2011 under controlled laboratory conditions. The results showed a small offset ranging from 0.08 to 0.23 ppm and a very small difference in calculated flux between the instruments.

The CO2 and CH4 fluxes were calculated from the changes in gas concentration as a function of time using linear fitting according to procedures by Crill et al. (1988), data has been corrected for the ambient air temperature and air pressure. As the replicate measurements (for each day) were all performed under stable and similar weather conditions we did not do any further corrections for environmental conditions. Release of gas from the ecosystem to the atmosphere is denoted by positive values and uptake by negative. Gas fluxes are expressed as mg m−2 h−1 of CH4 or CO2. Since we found a strong significant correlation (R = 0.967, p < 0.0001) between CH4 fluxes measured during light and dark measurement for individual plots, the mean of these two measurements were calculated and used in the CH4 flux calculations. NEE was the CO2 flux measured within the transparent chamber, while Reco were the dark measurement. GPP was calculated as the difference between NEE and Reco.

Pore-water analysis

Pore-water was sampled shortly after each gas flux measurement, for subsequent analysis of organic acids. Pore-water samples were drawn from stainless steel tubes (3 mm in diameter), which were permanently installed in 5, 10 and 15 cm below the peat surface. Five cm above the moss surface, each tube was closed by a three-way valve that enabled sampling without air penetration into the soil. From each plot a 9 ml mixed pore-water sample was drawn using a syringe, if it was not possible to retrieve 9 ml of sample, an equal amount from each depth was drawn. The sample was immediately filtered through a low protein binding non-pyrogenic sterile pre-rinsed filter (Acrodisc PF 0.8/0.2 μm diameter 32 mm) and frozen as soon as possible. Subsequently, the pore-water samples were analyzed for OA using a liquid chromatography-ionspray tandem mass spectrometry system. The system consisted of a Dionex (Sunnyvale, CA, USA) ICS-2500 liquid chromatography (LC) system and an Applied Biosystems (Foster City, CA, USA) 2000 Q-trap triple quadrupole mass spectrometer (MS). The LC–MS method and instrumental set-up is described in more detail in Ström et al. (2012). Due to the relatively high pH values in the mire the prevailing form of acetic acid is acetate (it is however termed as acetic acid throughout the paper).

Ecosystem variables

As a measure of the environmental conditions, in connection with each flux measurement and in close proximity to each plot replicate, we determined the water table depth (WtD, cm below moss surface), the active layer thickness (AL, cm below moss surface), photosynthetically active radiation (PAR) and soil temperature at 10 cm below surface (Ts), using a 150 mm digital temperature probe (Viking, Eskilstuna, Sweden). PAR (μmol m−2 s−1) was measured inside the chamber at 25 cm from the surface every minute, using a Minikin QTi data-logger (EMS Brno). Additionally, ambient incoming PAR and air temperature (Ta) were logged hourly (Minikin QTi) at 1 m from the ground surface throughout the growing season in 2011 and 2012. PAR and Ta data from 2010 were provided from the monitoring program ClimateBasis (Jensen and Rasch 2011). These, data were collected automatically about 1 km from the study site and measured at 2 m height. The WtD was measured at each block but to minimize disturbance not in individual plots. WtD was measured using a water permeable tube (2.5 cm in diameter) that was permanently installed in ground fitted with a float (made of cork). In 2011, the density and species composition of vascular plants was non-destructively estimated in each plot by counting the number of shoots of the three dominant vascular plant species.

Statistical analysis

To determine statistically significant differences between control, clipped and NV plots, with respect to gas fluxes, total OA and acetic acid concentration in pore water, Ts and AL, a general linear model repeated-measures analysis was performed (confidence interval adjustments were LSD). The model was made for each measuring year and took temporal development within seasons and the differences between blocks into account, as blocks and dates were random factors. Measurements earlier in the season with values around zero, due to an unproductive system, would render comparisons of the effects of the treatments on the carbon balance impossible. Therefore only fluxes measured after the onset of photosynthesis and with a negative NEE were included in the CO2 data analyses. Following the same reasoning, CH4 fluxes were included in the data treatment as long as the ecosystem was emitting measurable fluxes. The use of different time periods for CO2 and CH4 fluxes in 2012 should therefore be noted. Repeated-measures required gap filling of any data that were missing. Gap-filling was performed by computing a mean of the two measurements taken, on that particular plot, before and after the missing data point. Gap-filling was required in 2, 1.5 and 0 % cases for NEE, and Reco and CH4 respectively. Some OA data had to be removed due to contamination of the samples (one sampling date in 2010 and 2012 and two in 2011), in total 89.5 % of data was included in the data analyses. Gap filling was needed in 0.9 % of the cases for total OA and acetic acid.

To determine the variables that best could explain the plot scale variation in gas fluxes and any relationships between environmental variables a bivariate correlation (pearson 2-tailed test for significance) analysis was performed on the mean fluxes and ecosystem parameters (Ts, Ta, WtD, AL and PAR). To test the differences between the three vascular plant species in control and clipped plots data were first tested for normality and then an independent t test was performed.

Results

CH4 and CO2 fluxes

CH4

For each individual measurement year (2010–2012) a clear significant difference (p ≤ 0.001) between the CH4 fluxes in control and NV plots were seen. The flux was on average 84 % lower in NV than in control plots, with fluxes in NV plots ranging between 0.1–3.1, 0.0–4.6 and 0.0–3.5 mg CH4 m−2 h−1 for 2010, 2011 and 2012, respectively. Before the initiation of the increased grazing experiment (2 July 2010) the control plots and those that later were clipped were measured on four occasions (two times in August 2009 and two times in 2010). Here there were no significant differences (p ≥ 0.710) in CH4 flux between the plots (Fig. 1). Additionally, there was no significant difference (p = 0.571) in the mean CH4 flux between control and clipped plots after the initiation of simulated increased grazing in 2010. However, one year after the start of the experiment (2011), the mean CH4 flux was significantly lower in clipped than in control plots (p = 0.010) and this difference persisted during 2012 (p = 0.045). On average (2011 and 2012) the CH4 fluxes were 26 % lower in clipped plots compared to control plots (Fig. 1) and the difference was visible throughout the seasons (Fig. 7). The highest CH4 fluxes were measured around the 15 July (DOY 196–198). A late start of the growing season in 2012 is clearly seen in Fig. 7. That year the rapid increase of CH4 emission normally seen following the onset of the growing season started on DOY 190 while it the previous years started around DOY 170.

Fig. 1
figure 1

The mean measured CH4 fluxes (mg CH4 m−2 h−1 ± SE, 2009 n = 10, 2010 (before clipping experiment) n = 15, 2010 (after clipping experiment) n = 35, 2011 n = 60, 2012 n = 8 5) for control (stripped squares), clipped (open squares) and no vascular plants (NV) (closed squares) plots. Significant differences (repeated measures ANOVA) between control and treatment plots are indicated with asterisks above the bars, ***p ≤ 0.001, **p ≤ 0.01 and *p ≤ 0.05. Note the varying time periods between the years. The arrow indicates the start of clipping experiment on the 02/07/2010

Full size image

NEE

For each individual measurement year (2010–2012), a clear significant difference (p ≤ 0.001) between NEE in control and NV plots was seen (Fig. 2). On average NV plots were 113 % lower, with fluxes ranging between −52 to 218, −43 to 183 and −71 to 310 mg CO2 m−2 h−1 for 2010, 2011 and 2012, respectively. Before the initiation of the increased grazing experiment, there was no significant differences (p ≥ 0.724) in NEE between control and plots that later were clipped. However, after initiation of clipping there was an immediate and significant decrease in NEE which persisted throughout the years (Fig. 3; p = 0.003, p = 0.002 and p = 0.040 for 2010, 2011 and 2012, respectively). For all 3 years NEE was on average 35 % lower in clipped plots (Fig. 2a) compared to control plots. The lower uptake of CO2 in the clipped plots was mainly seen after the onset of the clipping each year. Consequently, the NEE values for the control and clipped plots were very similar before the first clipping of each season (Fig. 3a). In general, the mire acted as a net carbon sink for the main part of the measurement period in both control and clipped plots (NEE, Fig. 3a). 2011 was the least productive year with the lowest maximum uptake, while 2012 was the most productive year.

Fig. 2
figure 2

The mean measured NEE (a), Reco (b) and GPP (c) (mg CO2 m−2 h−1 ± SE, 2009 n = 10, 2010 (before clipping experiment) n = 15, 2010 (after clipping experiment) n = 35, 2011 n = 60, 2012 n = 60) for control, clipped and no vascular plants (NV) plots. Significant differences (repeated measures ANOVA) between control and treatment plots are indicated with asterisks above bars, ***p ≤ 0.001, **p ≤ 0.01 and *p ≤ 0.05. Note the varying time periods between the years. The arrow indicates the start of clipping experiment on the 02/07/2010

Full size image
Fig. 3
figure 3

NEE (a), Reco (b) and GPP (c) (mg CO2 m−2 h−1) for control and clipped plots from the whole measuring period in 2010, 2011 and 2012. The time is shown as day of year (DOY). Each data point is an average of each treatment (five plots). The clipping was performed in 2010 on DOY 183 and 198, in 2011 on DOY 196 and 213 (DOY) and in 2012 on DOY 201 and 226

Full size image

Reco

For each individual measurement year (2010–2012) a clear significant difference (p ≤ 0.001) between mean Reco in control and NV plots was seen (Fig. 2b). On average Reco in NV plots were 62 % lower, with fluxes ranging between 49–399, 27–465 and 0–217 mg CO2 m−2 h−1 for 2010, 2011 and 2012, respectively. Before initiation of the increased grazing experiment there were no significant differences (p ≥ 0.577) in Reco between control plots and those plots that were later clipped (Fig. 2b). Following the initiation of increased grazing there was a slight tendency towards lower Reco. This, however, was not significant and only seen in 2010 (p = 0.116, p = 0.222, p = 0.248 for 2010, 2011 and 2012, respectively). Maximum Reco was highest in 2010 and lowest in 2011. Again the late start of the growing season in 2012 is clearly seen in Fig. 3b.

GPP

For each individual measurement year (2010–2012) a clear significant difference (p ≤ 0.001) between GPP in control plots and NV plots were seen (Fig. 2c). The fluxes measured in NV plots was in general 89 % lower and ranged between 0–301, 0–465 and 0–234 mg CO2 m−2 h−1 for 2010, 2011 and 2012, respectively. Before initiation of the grazing experiment there were no significant differences (p ≥ 0.567) in GPP between control plots and those plots that later were clipped. Immediately after the start-up of the increased grazing simulation GPP decreased in clipped plots. The lower GPP was consistent throughout the years (p = 0.009, p = 0.015 and p = 0.019 for 2010, 2011 and 2012, respectively). The GPP in clipped plots were on average 21 % lower than control plots for all three years (Fig. 2c). The difference in GPP was consistent through the main part of the measurement periods (Fig. 3c). 2011 had consistently lower GPP than 2010 and 2012, while these years had very similar GPP ranges. Again the late start of the growing season in 2012 (DOY 190 compared to DOY 170 the previous years) is clearly seen in Fig. 3c.

Pore-water chemistry

The OA pool was dominated by acetic acid and additionally included citric acid, formic, glycolic, lactic, malic, oxalic, succinic, and tartaric (Table 1). The concentration of OA ranged between, 48–2796, 9–2590 and 17–524 μg C l−1 in control plots, 16–3652, 40–2569 and 25–726 μg C l−1 in clipped plots and 28–2888, 30–2579 and 15–537 μg C l−1 in NV plots, for 2010, 2011 and 2012, respectively. A strong significant correlation between OA and acetic acid was found (R = 0.990, p < 0.0001). The concentrations of acetic acid measured in NV plots ranged between 0 ≥ 2402, 4.1 ≥ 2402 and 2.7–342.6 for 2010, 2011 and 2012, respectively.

Table 1 The percentile distribution of the organic acid for the three measuring years for control, clipped and plots with no vascular plants (NV)
Full size table

No significant difference was found in the total OA concentration between control and treatment plots for any of the measurement years (Fig. 4a) (clipped plots p = 0.759, p = 0.119 and p = 0.227, and NV plots p = 0.943, p = 0.105 and p = 0.643 for 2010, 2011 and 2012, respectively). However in 2012, the third year after the experiment was initiated, a significantly (p = 0.044) lower (−27 %) mean acetic acid concentration was found in clipped plots compared to control plots and a strong tendency for NV plots (p = 0.066) was also found (Fig. 4b). The first 2 years no significant differences were seen in the acetic acid concentration between treatments (clipped plots p = 0.929, p = 0.136 and NV plots p = 0.767, p = 0.106 for 2010 and 2011, respectively). The difference between treatments in acetic acid concentration tended to decline later in the season and was, irrespective of treatment and year substantially higher in the beginning of the growing season (Fig. 5).

Fig. 4
figure 4

The mean concentration (μg C l−1 ± SE, 2010 n = 45, 2011 n = 60, 2012 n = 65) of organic acid (a) and acetic acid (b) in control, clipped and plots with no vascular plants (NV). Significant differences (repeated measures ANOVA) between control and treatment plots are indicated with asterisks above the bars, ***p ≤ 0.001, **p ≤ 0.01 and *p ≤ 0.05. Note the varying time periods between the years

Full size image
Fig. 5
figure 5

The measured acetic acid concentration (μg C l−1) in control and clipped plots, for the whole measuring period in 2010, 2011 and 2012. The time is shown as day of year (DOY). Each data point is an average of five plots. The clipping was performed in 2010 on DOY 183 and 198, in 2011 on DOY 196 and 213 (DOY) and in 2012 on DOY 201 and 226

Full size image

Species composition

During the vegetation survey, the main vascular plant species within the plots were Carex stans 42 ± 2.5 (% ± SE), Dupontia psilosantha 32.7 ± 2.1 and Eriophorum scheuchzeri 24.3 ± 2.6. There was no significant effect of increased grazing on the vegetation composition after 1 year of grazing. However a strong tendency (p = 0.07) towards a higher total number of vascular plants in the control plots were seen (Fig. 6). Additionally, the number of Eriophorum shoots tended to be lower (p = 0.096) in clipped compared to control plots. For Carex and Dupontia the differences were less pronounced (p = 0.231 and 0.589, respectively, Fig. 6).

Fig. 6
figure 6

Vegetation analyses from the experiment plots in NE Greenland, analyses are made in 2011. The dominating species in the plots were: Carex stans, Dupontia psilosantha and Eriophorum scheuchzeri. The figure shows the mean number of tiller of the three dominating vascular plant species and the total number of tillers ± SE, for control and clipped plots. No significant differences between the controlled and clipped plots for any of the species

Full size image

Environmental variables

Since all plots were situated in close proximity and in the same vegetation type there was no difference between blocks or plots in Ta, PAR, AL, peat layer depth or WtD. In 2010 Ts at 8–10 cm was however significantly lower (mean 0.7 °C) in NV plots compared to control and clipped plots (p = 0.031). In 2011 and 2012 no significant differences were observed p = 0.280 and p = 0.444 for 2011 and 2012, respectively. There was no significant difference in Ts (p ≥ 0.103) between control and clipped plots for any of the years. The date of snow melt differed between the years, in 2012 it melted more than 10 days later than the previous years (Table 2), due to the large amount of snow that fell in the beginning of 2012. Since WtD mainly depends on the amount of snow precipitation, the level decreases over the summer. In 2012 the maximum WtD was not reached until the end of autumn, which correspondences with the high amount of water input from melting snow that summer. No significant differences (p ≥ 0.550) between the mean measured AL were found for the three treatments.

Table 2 Measured ecosystem parameters for the different measuring periods in the mire in NE Greenland, Zackenberg
Full size table

Comparing the monthly mean environmental variables for 2010, 2011 and 2012 we found a significant correlation between PAR and Ts (R = 0.888, p = 0.001) and between WtD and AL (R = 0.780, p = 0.013). There were, however, no significant correlations between any of the other measured environmental variables; Ta and Ts (R = 0.489, p = 0.182), Ta and PAR (R = 0.296, p = 0.440), Ta and WtD (R = −0.162, p = 0.678), Ta and AL (R = −0.232, p = 0.548), WtD and Ts (R = −0.285, p = 0.457), WtD and PAR (R = 0.129, p = 0.741), AL and Ts (R = −0.144, p = 0.713) or Al and PAR (R = 0.100, p = 0.798).

Controls of fluxes

In order to identify the likely controls of fluxes in the plots, a correlation analysis between the measured mean fluxes, OAs and environmental soil variables was performed.

The analysis showed highly significant correlations between all flux measurements; CH4 and NEE (R = −0.926, p ≤ 0.001), CH4 and Reco (R = 0.957, p ≤ 0.001), CH4 and GPP (R = −0.976, p ≤ 0.001) and NEE and Reco (R = −0.868, p = 0.002). No correlation was found between gas fluxes and OA or acetic acid concentration, OA and CH4 (R = 0.322, p = 0.399), OA and NEE (R = −0.157, p = 0.688), OA and Reco (R = 0.528, p = 0.144), acetic acid and CH4 (R = 0.385, p = 0.306), acetic acid and NEE (R = −0.192, p = 0.621), acetic acid and Reco (R = 0.580, p = 0.101). No significant correlations between the seasonal mean CO2 and CH4 fluxes and the measured soil properties (AL and Ts) (p ≥ 0.4) were found.

Discussion

The magnitude of CO2 and CH4 fluxes and the patterns during the growing season varied between the three measurement years (Figs. 1, 2, 3 and 7). The differences between the mean values (CH4, NEE, GPP and Reco, Figs. 1 and 2) for the three years is largely due to the different length of the measurement period each season and to the inter-annual variation in WtD. The measured flux magnitudes during the measurement periods were consistent with fluxes previously measured in the same mire (Mastepanov et al. 2008; Ström et al. 2012).

Fig. 7
figure 7

CH4 fluxes (mg CH4 m−2 h−1) for control and clipped plots from the whole measuring period in 2010, 2011 and 2012. The time is shown as day of year (DOY). Each data point is an average of each treatment (five plots). The clipping was performed in 2010 on DOY 183 and 198, in 2011 on DOY 196 and 213 (DOY) and in 2012 on DOY 201 and 226

Full size image

Grazing effects on CO2 fluxes

The magnitude of the ecosystems responses to herbivory depends on type of ecosystem as well as grazing pressure (Mulder 1999). However, in many studies conducted in the Arctic have reported the same trends as found in this study, with a decrease in NEE with grazing (Cahoon et al. 2012; Elliott and Henry 2011; Sjögersten et al. 2008, 2011; Van der Wal et al. 2007). Over the 3 years we found a 35 % reduction in mean seasonal NEE in clipped plots compared to control plots (Fig. 2). The reduction can most likely be explained by the removal of photosynthetically active biomass and hereby less carbon uptake, as studies have shown a strong relationship between NEE, GEP (gross ecosystem photosynthesis) and the living plant biomass (Sjögersten et al. 2008; Ström and Christensen 2007). Additionally, the large differences in NEE just after the plots have been cut (Fig. 3) supports the hypothesis that reduced biomass leads to reduced CO2 uptake. In a study from west Greenland, grazing induced a decrease in NEE by 190 %, and a change in the vegetation composition from being graminoid dominated when grazed to becoming shrub dominated when un-grazed (Cahoon et al. 2012). A vegetation alteration between graminoids and shrubs will result in a much larger NEE decrease and therefor their percentile is not fully comparable to our findings. NEE was 113 % lower in our NV plots, which were expected as most of the photosynthetic biomass (all vascular plants) was removed.

Sjögersten et al. (2011) and Van der Wal et al. (2007) found a strong reduction in above-ground biomass, below-ground biomass and C storage following grazing. Their studies were made in areas dominated by geese that are both grazing and grubbing, which may disturb the ecosystem hugely. In some cases it leads to exposed organic layer, which can result in erosion and huge losses of carbon. Speed et al. (2010) estimated that goose grazing and grubbing in some areas of Svalbard could result in a 75 % loss of carbon from vegetation and organic soil pools. The muskoxen graze the vegetation a couple of centimeters above the surface in the mire (Kristensen et al. 2011), which is much less invasive for the vegetation than geese grazing. It is therefore most likely that the below-ground biomass and the carbon stock are less disturbed in areas dominated by large herbivores than by geese. On the other hand a factor that is not considered in this study is the loss of accumulated CO2 and CH4 from the mire due to trampling, which may have substantial impact on the vegetation and pattern of gas flux emissions, but not directly on its production.

The much lowered respiration (mean of −62 % for all years, Fig. 2) found in plots with no vascular plants was expected, as the autotrophic respiration should be strongly reduced following removal of both above- and below-ground vascular vegetation. Several studies on arctic ecosystems have found a decrease in respiration with herbivory (Cahoon et al. 2012; Sjögersten et al. 2011; Stark and Grellmann 2002; Van der Wal et al. 2007). A likely explanation to the often observed reduction in Reco may be that graminoids in the arctic are able to reduce their carbon allocation to below-ground structures such as roots when grazed (Chapin 1980). This would most likely lead to lower Reco, as the root density and the below-ground carbon concentration in general are related to respiration from the soil (Chapin and Ruess 2001; Hanson et al. 2000; Hogberg et al. 2001). In a study by Richards (1984) defoliation resulted in a 50 % reduction in root growth (Richards 1984). Our study could however not confirm these findings since we found no significant differences in Reco between control and clipped plots, although a small but not significant tendency towards a reduction in Reco was observed in 2010 (Fig. 2b). In wet arctic mires CO2 respiration is to a high degree controlled and reduced by high WtD; this may explain the differences seen in this study in comparison to other studies that are performed in drier habitats.

Grazing effects on methane emissions

Many previous studies have shown a correlation between CH4 emissions and net ecosystem production (NEP) or GPP (Bubier 1995; Christensen et al. 2000; Joabsson and Christensen 2001; Ström and Christensen 2007; Waddington et al. 1996). The commonly suggested explanation for these relationships is the effect of productivity and photosynthesis on methanogenesis, as a higher carbon uptake would result in more supply of methanogenic substrates to the root zone and subsequently to higher CH4 emissions (Joabsson et al. 1999). Offering support to these studies we found a strong correlation between ecosystem productivity (NEE/NEP and GPP) and CH4 emission. One might hypothesis that a lower CH4 emission could be expected from a grazed ecosystem, since grazing is found to decrease GPP and NEP. Indeed, our study supports this hypothesis as we found that CH4 emission was 26 % lower in clipped plots compared to control plots (Fig. 1). This contrast the only other study (to our knowledge) that is made on herbivory and CH4 fluxes in an wet high arctic habitat, that reported no changes in the CH4 fluxes between grazed and un-grazed plots (Sjögersten et al. 2011). Their study was, based on a 4 years old exclosure experiment in an area with geese and very low CH4 fluxes (−0.046 to 0.025 mg CH4 m−2 h−1), and CH4 fluxes were only measured a few times over one growing season. Consequently, responses on CH4 fluxes may not be expected to the same extent as in our productive mire.

The large differences between plots that were clipped and NV plots, indicates the importance of the presence of vascular plants, which are known to mediate methane transport directly from anoxic peat depth to the atmosphere, thereby decreasing methane oxidation in oxic upper peat layers (Bellisario et al. 1999; Frenzel and Karofeld 2000; Greenup et al. 2000; King et al. 1998; Schimel 1995; Whalen 2005). Whalen (2005) showed that the absence of vascular plants could reduce CH4 emission by 50–85 %, which is in line with our findings. In particular, Eriophorum species are often mentioned as being important for methane emissions (Frenzel and Rudolph 1998; Greenup et al. 2000; Joabsson and Christensen 2001; Schimel 1995; Ström and Christensen 2007; Ström et al. 2005, 2012). Offering some further support to the importance of vascular plant presence and composition to the CH4 flux, we found a strong tendency towards a higher total number of vascular plant in the control plots. Additionally, the number of Eriophorum shoots tended to be lower (Fig. 6) in clipped compared to control plots. See below for further discussions on grazing effects on vegetation composition.

Grazing effects on plant community

In a study by Ström et al. (2012) a strong linkage between density of Eriophorum tillers, acetic acid concentration in pore water and a CH4 emission was demonstrated. As mentioned above, we found a tendency towards a higher number of Eriophorum shoots in control plots (Fig. 6). A linkage between high CH4 fluxes, high acetic acid concentration in the control plots, and a higher number of Eriophorum shoots corresponds well to the findings by Ström et al. (2012). The tendency towards a lower number of Eriophorum shoots in clipped plots may indicate that this species is more sensitive to higher grazing pressure than others. Additionally we found a strong tendency (p = 0.07) towards a higher total number of vascular plant in the control plots (Fig. 6). It might be speculated that the short arctic summer may also be too short for plants to fully recover from the loss of above ground tissue (Elliott and Henry 2011), which makes these ecosystems more sensitive towards grazing than others. The vegetation survey in our study was, however, performed already 1 year after the start of simulated grazing experiment and it is not unlikely that the vegetation composition has continued to change in clipped plots. Further studies are therefore necessary to confirm all the speculations above.

Grazing effects on substrate availability

The pore-water concentration of easily available substrates for methanogens and the magnitude of CH4 emissions are closely linked (Christensen et al. 2003; Ström and Christensen 2007; Ström et al. 2005, 2012). The dominating organic substrate for methane production in wetlands is acetic acid (Ström and Christensen 2007; Ström et al. 2003, 2005, 2012), which alone have been found to account for 3.9 % of DOC (Ström et al. 2012). Not many studies on how grazing influence substrate availability has been conducted over longer periods of time and at in situ field conditions. A limited number of short time studies have, however, shown an increase in root exudation shortly after the vegetation has been “grazed” (Butenschoen et al. 2008; Hamilton Iii et al. 2008; Paterson et al. 2005). In contrast to these studies we found that after 3 years of increased grazing pressure substrate availability decreased with 27 % (Fig. 4b), thus stressing the importance of long-term studies before any generalization considering ecosystem responses to grazing can be made. Herbivory-induced changes in the growth pattern of sedges and graminoids include increased shoot production, increased leaf length and more leaves per shoot (Beaulieu et al. 1996; Chapin 1980). Graminoids generally preserve a larger reserve of below-ground carbon and nutrients than woody plants, which supports the fast regrowth of shoots after they have been grazed (Bryant et al. 1983; Chapin 1980; Green and Detling 2000; Mulder 1999). In a study by Chapin (1980) a decrease in root growth was found with defoliation of Eriophorum and Carex. In a high arctic area with high goose grazing pressure, Dupontia fisheri and Eriophorum Scheuchzeri were able to produce new shoots on the expense of below-ground reserves (Beaulieu et al. 1996). This is consistent with our findings of less available substrate in form of acetic acid in clipped plots. Subsequently, the equally low concentrations of acetic acid in NV plots are not surprising since the main producers of “fuel” for acetic acid production in the root zone are removed. It is important to mention that we in our study cannot discriminate between acetic acid or organic acids in general that are produced by fermentation or by root exudation, as our samples are taken from the peat layer where both processes are active.

Climate warming and grazing

Climate warming is proceeding faster in the Arctic than elsewhere on Earth (ACIA 2005). One of the evidence and predictions for a warmer Arctic is a greening and a transformation in plant composition towards a higher density of shrubs (Hill and Henry 2011; Hudson and Henry 2009; Myers-Smith et al. 2011; Tagesson et al. 2010). Several studies with combined warming and grazing have showed that grazing in the Arctic can have a larger effect on the carbon uptake than higher temperatures. Grazing by vertebrate herbivores may mitigate the increase in biomass production or the change toward more shrubs that predicted and observed with increasing temperatures in the Arctic (Cahoon et al. 2012; Klein et al. 2007; Olofsson et al. 2009; Post and Pedersen 2008; Rinnan et al. 2009; Sjögersten et al. 2008). Though an expansion of shrubs often is found in response to warming there is currently no indications of such a trend in the Zackenberg area (Schmidt et al. 2012), which is likely due to high density of muskoxen in the area (Myers-Smith et al. 2011). This delicate interplay between muskoxen and shrubs (Post and Pedersen 2008) makes muskoxen an important component for the function of this high arctic mire. During the last decade the moss density in the Zackenberg mire has increased and the ecosystem has become drier (Schmidt et al. 2012). Our study plots with no vascular plants had a much higher density of mosses and additionally much lower methane emission compared to control plots.

Conclusion

In our study we found that increased grazing in a high arctic mire plays a significant role for the carbon cycle and the CH4 emission. We found that net ecosystem uptake decreased immediately after the initiation of the clipping experiment, for the following 3 years the total mean decrease was 35 %. The third year into the clipping experiment significantly lower substrate (acetic acid) availability for CH4 production was found, the concentration was 27 % lower that year. This reduction indicates that these vascular plants, when exposed to a high grazing pressure, to a higher degree allocate the carbon to above-ground regrowth then to the root system. The second year into the experiment a vegetation analysis showed a strong tendency (p = 0.07) that clipped plots had lower number of vascular plants. The lower substrate availability, number of vascular plants and together with a decrease in CO2 uptake explain the significant reduction we found in CH4 emission the second year into the experiment, at the clipped plots. The following 2 years the mean decrease in CH4 emission was 26 %. If the grazing pressure becomes so high that mosses become by far the most dominating vegetation type, we can expect a large decrease of CO2 uptake and CH4 emission. In our plots with no vascular plants (NV) we found a reduction in CO2 uptake and CH4 emission by 113 and 84 %, respectively. This study shows what an important role herbivores play in the ecosystem and that their influence should not be overlooked, when discussing the carbon balance of the arctic.