Drought alters carbon fluxes in alpine snowbed ecosystems through contrasting impacts on graminoids and forbs

Authors


Author for correspondence:
David Johnson
Tel: +44 1224 273857
Email: d.johnson@abdn.ac.uk

Summary

  • Climate change is predicted to increase the frequency of drought events in alpine ecosystems with the potential to affect carbon turnover.
  • We removed intact turfs from a Nardus stricta alpine snowbed community and subjected half of them to two drought events of 8 d duration under controlled conditions. Leachate dissolved organic carbon (DOC) was measured throughout the 6 wk study period, and a 13CO2 pulse enabled quantification of fluxes of recent assimilate into shoots, roots and leachate and ecosystem CO2 exchange.
  • The amount of DOC in leachate from droughted cores was 62% less than in controls. Drought reduced graminoid biomass, increased forb biomass, had no effect on bryophytes, and led to an overall decrease in total above-ground biomass compared with controls. Net CO2 exchange, gross photosynthesis and the amount of 13CO2 fixed were all significantly less in droughted turfs. These turfs also retained proportionally more 13C in shoots, allocated less 13C to roots, and the amount of dissolved organic 13C recovered in leachate was 57% less than in controls.
  • Our data show that drought events can have significant impacts on ecosystem carbon fluxes, and that the principal mechanism behind this is probably changes in the relative abundance of forbs and grasses.

Introduction

Water availability is a key factor that limits plant growth and the activity of soil microorganisms that regulate biogeochemical cycles. On the one hand, water limitation has been demonstrated to have severe effects on plant productivity, reproduction and a range of important ecosystem functions. On the other hand, too much water can also have a great effect and so plants have evolved a range of strategies to deal with consistently high soil water potentials. It is of particular concern, therefore, that the climate in north temperate ecosystems is predicted to become more extreme, with an increased frequency of drought events being forecast (IPCC, 2007). In Scotland, short-term extreme drought events are predicted to become more frequent under Intergovernmental Panel on Climate Change (IPCC) scenarios (Vidal & Wade, 2009). These events are likely to be a consequence both of increases in air temperatures and changes in precipitation patterns. Although winter precipitation is expected to increase, snow fall, and therefore snow cover, is predicted to decrease, leaving alpine and upland areas exposed to more freeze–thaw cycles and reducing water availability from snow-melt during the summer months. Climate change has the potential to affect all natural processes, since temperature and water availability are among the main factors in determining ecosystem functioning (Aerts, 2006). However, ecosystem responses may be variable and climate manipulation studies are needed to understand and predict how ecosystems might respond to these changes (Shaver et al., 2000). Identifying how alpine and upland ecosystems respond to drought events is therefore crucial if we are to fully understand the consequences of man-made climatic change in these areas.

Alpine ecosystems are frequently dominated by slow-growing plant species adapted to low nutrient availability and stable but extreme climatic conditions. These characteristics make them vulnerable to large shifts in air temperature and precipitation (Tinner & Kaltenrieder, 2005; Vittoz et al., 2009), with the potential to jeopardize carbon (C) sequestration, water quality and conservation of biodiversity. In the UK, the low- to mid- alpine zone occupies c. 3% of the land surface and contains the most extensive remaining areas of near-natural plant communities (Thompson & Brown, 1992). The largest and best preserved areas of alpine vegetation occur in Scotland, while smaller areas are present in England and Wales. The alpine zone occurs at a lower altitude in the UK compared with continental Europe, as a result of the UK’s oceanic climate which causes high wind speeds and rainfall. British alpine plant communities contain elements of both Arctic and alpine floras and are strongly influenced by the oceanic climate. This combination of characters results in a number of plant communities that are either unique or particularly well represented in the UK (Thompson & Brown, 1992). In Scotland, alpine ecosystems vary from exposed fell-field communities on shallow, wind-blown soils, to a range of early and late-melting snowbed communities in more protected situations on deeper highly organic soils. The large stores of organic matter in early-melt locations makes these snowbed communities of particular interest from a C cycling perspective. Moderate drought stress could stimulate loss of C by reducing water content and increasing oxygen availability while more extreme drought events could lead to water limitation and reduce productivity and below-ground C allocation.

Although the response of alpine vegetation to drought is little studied, a range of impacts have been seen in other semi-natural, low productivity systems that may be analogous to alpine systems. For example, in heathland, removal of precipitation for 2 months during the summer had contrasting effects depending on the natural water availability (Jensen et al., 2003). In a drier site, soil CO2 efflux was reduced while the opposite was the case in a wetter site. In boreal peatland, dissolved organic carbon (DOC) concentrations were 25% greater under natural drought conditions, but the effects were buffered by subsequent snowmelt events (Jager et al., 2009). Other studies have demonstrated interactions between DOC loss and natural drought phenomena and suggest a key role of phenol oxidases in regulating C fluxes in leachates (Freeman et al., 2004b; Toberman et al., 2008). The impacts of drought on below-ground C flux are particularly important because it is the supply of labile C, primarily comprising recent plant assimilate, that provides energy for free-living saprotrophic microorganisms in the rhizosphere. Moreover, there is evidence that ‘priming’ can occur whereby inputs of new C stimulate the breakdown of older, more recalcitrant forms (Fontaine et al., 2007) that primarily dictate the long-term storage pools in soils. Identifying the importance and quantifying the flux of recent plant assimilate below ground is possible through the application of isotope tracer techniques. For example, in Sphagnum-dominated peatland, application of 13CO2 tracers demonstrated that primary productivity induced by elevated CO2 increased 10-fold the recovery of recent assimilate in leachate (Freeman et al., 2004a). Transfer of C to leachate is important because, first, a component of this pool of C is thought to turn over very quickly, thus providing energy for important processes such as nutrient mineralization; secondly, DOC in leachate is easily transported, enabling C to be redistributed over large spatial scales or lost from ecosystems (Holden, 2005); and thirdly, DOC derived from leachates in upland areas reduces water quality and is a major economic problem for the water supply industry (Mitchell, 1991).

The uncertainty in how alpine plant communities and the ecosystems they support will respond to drought stress indicates a need to undertake manipulative experiments to fill this gap in our understanding. This is important from the perspective of understanding how plant communities respond to drought stress and the subsequent consequences for key ecosystem processes, such as C turnover and storage. By gaining an insight into the pathways and mechanisms regulating C fluxes, it will become easier to predict how climate variability will impact on north temperate alpine ecosystems. In this paper, therefore, we report an experiment designed to quantify the major fluxes and pools of C derived from recent plant assimilate in an alpine snowbed community in response to drought stress. We used intact vegetation monoliths transported to controlled-environment growth rooms to test the effects of simulated summer drought events on C fluxes into leachate and ecosystem respiration, and on the productivity of the three main plant functional types found in snowbeds. The use of 13CO2 pulse-labelling enabled us to test the hypothesis that drought would affect the allocation of recent plant assimilate into vegetation biomass, bulk soil, ecosystem respiration and leachate. We predicted that drought would increase CO2 flux from snowbeds because the resulting fluctuations in the water content of the soils would stimulate the productivity of the microbial community (Jensen et al., 2003). In addition, we predict that DOC flux in leachate would also increase in response to drought as seen in other correlative studies in seminatural ecosystems (Jager et al., 2009). Given that the communities have adapted for millennia to the prevailing climate, we also predicted that drought events imposed at a modest intensity would not have a major direct impact on the physiology and biomass of the plants.

Materials and Methods

Sampling and experimental design

This experiment used soil cores collected on 12 May 2009 from a Nardus stricta L. snowbed community located on a south-facing slope of Culardoch, Cairngorms, Scotland (latitude, 57.07331°N; longitude, 3.34002°W). The dominant plant species was N. stricta, but the community also contained Vaccinium myrtillus L., Deschampsia flexuosa (L.) Trin., Potentilla erecta (L.) Raüschel and Sphagnum spp. The soil is highly organic (mean % C = 42.1%, mean % N = 1.3%) and there was little apparent profile development. Intact turfs, 15 cm in diameter, were removed using a custom-made stainless steel soil corer to a depth of 30 cm, which was designed to prevent compaction of cores. The cores were sampled throughout the snowbed and removed in pairs based on approximate similarity of the plant communities. The cores were immediately placed into close-fitting PVC pipes, trimmed to 20 cm depth, and transported to the Macaulay Land Use Research Institute on the same day. A total of 20 cores were removed and the bases were covered with a coarse nylon mesh. They were maintained outdoors for 12 d and then moved into a controlled-environment growth room (18: 6 h/15 : 10°C day : night – these are close to the summer temperatures recorded at the field site; photosynthetically active radiation (PAR) = 300 μmol m−2 s−1) on 25 May. The cores were split into two groups of 10 that were treated identically up to the start of the 13CO2 labelling experiment. Five of the cores from each group were assigned to the drought treatment. All 20 cores were arranged randomly in the growth room and mounted on a 20-cm-diameter funnel suspended over a second 20-cm-long PVC pipe. This permitted us to collect all leachate from each core for the duration of the experiment. Soil temperature was monitored every hour in three cores from the controls and three from the drought treatment, and soil moisture was monitored in one control core and one droughted core with a theta probe.

Each core received 250 ml of artificial rainwater, applied every weekday, using a recipe relevant to northwest Scotland (UK Review Group on Acid Rain, 1997). This contained: CaCl2 2H2O (7.5 μM l−1), KH2PO4 (3.5 μM l−1), Na2SO4 (17.5 μM l−1), MgCl 6H2O (20 μM l−1), KCl (4 μM l−1), NaCl (45 μM l−1), NH4NO3 (20 μM l−1) and was adjusted to pH 4.9. Two drought treatments were imposed on the cores, which involved stopping the addition of water for two periods each of 8 d duration (15–23 June and 6–14 July). This had the effect of reducing soil moisture by a maximum of 11.9% during the first event and a maximum of 23.7% in the second event; soil moisture returned to control values c. 2 wk after the end of the drought period. Leachates from the group of 10 cores that were not assigned to the 13CO2 labelling were collected every weekday morning from 12 June to 20 July 2009, the volumes measured and immediately frozen. All leachates were sampled for total C by UV/persulphate oxidation.

13CO2 pulse labelling

On 20 July, the 10 cores not assigned to the 13CO2 labelling were removed and stored in the cold room. These were used to determine natural abundance δ13C isotopic signatures of roots, shoots and soil. On 21 July, the vegetation in the remaining group of 10 cores was pulse-labelled with 13CO2 for 4 h from 11:00 to 15:00 h. Clear plastic bags that enclosed the entire plant communities (c. 2.5 l volume) were taped to each core and an inlet tube inserted into the air space from below. The tubes were connected, via a series of flow meters and valves, to a 7000 l tank of compressed air containing 99 atom%13CO2 at 370 ppm (Spectra Gases Ltd, Cambridge, UK). The gas was supplied to each core using tank pressure to maintain a flow rate of 0.2–0.3 l min−1, which was sufficient to provide slight positive pressure in the bags but resulted in a dynamic flowthrough labelling procedure. Preliminary measurements also showed that this flow rate was sufficient to maintain ambient CO2 concentrations in the canopy headspace for the duration of the labelling. After labelling, a subsample of the green tissue from graminoids, forbs and bryophytes was immediately removed from each core and frozen. The pulse-chase period was 140 h after which the cores were removed and destructively harvested.

Sampling and analysis of leachates, soil, vegetation and CO2 fluxes

During the labelling study, leachates were sampled on the 22 (1 d post-labelling), 23, 24 and 27 July. In addition to total C, the δ13C signatures were determined by liquid oxidation as described by Potthoff et al. (2003). The resulting CO2 was analysed using a Thermo Finnigan Delta PlusXP isotope ratio mass spectrometer (IRMS) interfaced to a Gas Bench II and PAL autosampler (Thermo Finnigan, Bremen, Germany).

Ecosystem CO2 fluxes were determined on 22 and 24 July. Net CO2 exchange was measured at PAR 300 μmol m−2 s−1 by sealing a 2.5 l volume Perspex chamber over the cores. The chamber had a battery-operated fan inside and a suba-seal gas sampling port and two inlet lines. The gas lines were connected to a PP System EGM 4 infra-red gas analyser (IRGA). Once the chamber was closed, CO2 concentrations were monitored every 40 s for 440 s (to give eight readings). To measure ecosystem respiration, the procedure was repeated in the dark except that the sampling frequency was once every 20 s for 220 s. At the start and end of the monitoring period, a 25 ml sample was injected into evacuated glass vials (Exetainers, Labco Ltd., High Wycombe, UK) for subsequent isotopic analysis, as previously described. Linear regressions were obtained from the CO2 data and gross photosynthesis was determined as the sum of these two measurements. A further set of gas samples was obtained for isotopic analysis on 27 July.

At the end of the experiment, the above-ground biomass of forbs, graminoids and bryophytes was removed, freeze-dried and weighed. For C analysis, 1 mg of each ground plant biomass sample (natural abundance controls, immediately post-labelling and final harvest) or bulk soil (1/8th of the core removed and sieved) was weighed into a 4 × 6 mm tin cup and analysed for total C content and δ13C by continuous-flow IRMS using a Europa Scientific ANCA-NT stable isotope analyser with ANCA-NT Solid/Liquid Preparation Module (Europa Scientific Ltd, Crewe, UK). Root samples were extracted from 1/8th of the core volume and treated similarly.

All 13C data were converted from δ13C values (‰) to 13C excess values (atom%) using the following equations:

To determine the isotopic ratio 13C : 12C of samples (Rsample):

image

where 0.011237 is the ratio of 13C : 12C in the Pee Dee Belemnite standard.

To determine the 13C abundance (atom%) of samples:

image

The atom% values were converted to 13C atom%‘excess’ by subtracting the %13C of unlabelled controls (i.e. samples not exposed to enriched 13C) from each enriched sample. To calculate the total amount of 13C derived from the pulse of 13C, the atom% excess was integrated with the total C content found in that sample (e.g. DOC in leachate, C content of roots, shoots and soil). Respired 13C was calculated by multiplying 13C atom% excess figures by the corresponding values of total C flux. The quantity of 13C fixed by each plant functional type during the pulse was calculated from the 13C contents of shoot subsamples taken immediately following labelling and integrating these data with the final C content of total biomass of leaf tissue obtained at the end of the experiment. 13C in other pools was calculated as a percentage of this initial 13C uptake.

Statistical analyses

Data were analysed in Minitab 15 (Minitab Ltd., Coventry, UK) using the General Linear Model (GLM) function. Most of the analyses required that the statistical model had a split-plot structure, to account for analysis of different sample types from the same core (such as the biomass of forbs, graminoids, bryophytes) or for time course measurements of samples (such as DOC) from the same core, which cannot be treated as independent measurements. Fixed factors were therefore treatment (i.e. drought or control), core identity and either time or vegetation type (i.e. forbs, graminoids and bryophytes); these latter two factors were nested within ‘core identity’. All residual variances were checked for homogeneity and data transformed where necessary.

Results

The two drought events resulted in cessation of leachate draining from the cores during each 8 d period (Fig. 1a). Thereafter, the volume of leachate collected from the droughted cores increased to control values within 5 d after cessation of the last simulated drought. At the start of the pulse labelling, the volume of leachate collected was slightly greater in the control cores but the volumes converged throughout the following 5 d (Fig. 1a). The flux of DOC from the group of 10 cores (that were not labelled with 13CO2) between 12 June and 20 July was variable (Fig. 1b) and ranged from 0.65 to 2.7 mg C d−1 in the control cores. The mean total amount of DOC (± SE) recovered from the controls was 52.6 ± 8.3 mg C, whereas it was only 20.0 ± 10.9 mg C in the drought treatment; these values were significantly different (= 0.001; F1,9 = 28.6). Although a large part of this effect was the result of the lack of any leachate being produced in the treated cores during the drought periods, a similar effect was seen when only DOC recovered in the nondrought periods was analysed. Here, 29.4 ± 4.6 mg C was recovered from the controls and 15.9 ± 8.6 mg C was recovered from the droughted cores (= 0.015; F1,9 = 9.5).

Figure 1.

Volume of leachate collected (a) and flux of dissolved organic carbon (DOC) (b) from control (closed circles) and droughted (open circles) cores removed intact from a Nardus stricta snowbed (± SEM). Half of the cores were subjected to two simulated droughts (grey panels) of 8 d duration during which time they received no water. Otherwise, each core received 250 ml of artificial rainwater every weekday. The dashed line represents the start of the 13CO2 pulse label.

The above-ground biomass at the end of the experiment revealed differences in the abundance of the dominant vegetation types, namely forbs and graminoids (Fig. 2), that was reflected by an overall significant functional type × treatment interaction (F2,29 = 6.7; = 0.008). In the droughted cores, forb biomass was 68% greater than in the controls (F1,9 = 5.5; = 0.047) while the opposite pattern was seen in graminoid biomass, which was 69% less in the droughted treatment (F1,9 = 6.03; = 0.040). Bryophyte biomass did not differ significantly (= 0.06) between treatments, although there was tendency for increased biomass in the droughted cores. The total above-ground biomass in droughted cores was 3.5 ± 0.7 g DW per core compared with 4.2 ± 0.2 g DW per core in the controls, but this was not significantly different. However, this appeared to be driven by a single droughted core which had greater biomass than the rest (6.0 g DW compared with a range of 2.1–3.8 g DW the other four cores). When this core was removed from analysis, mean total above-ground biomass was 2.9 g DW per core in the droughted treatment compared with 4.2 g DW per core in the controls (F1,8 = 10.2; = 0.015). We nevertheless retained this core in all future analyses.

Figure 2.

Above-ground biomass of forbs, graminoids, bryophytes and the total recovered from control (open bars) and droughted (hatched bars) cores at the end of the experiment (± SEM). Asterisks indicate significant differences within a vegetation type (< 0.05); ns, nonsignificant.

The above-ground biomass reflected the amount of 13C fixed from the 13CO2 pulse (Fig. 3a). In the control cores, the mean total amount of 13C fixed by all plants was 10218 ± 3346 μg and in the droughted cores it was 7232 ± 3348 μg. There was a marginal functional type × drought interaction (F2,29 = 3.01; = 0.078), with graminoids tending to fix less 13C (4377 ± 2131 μg) in the droughted cores than in the controls (8662 ± 2863 μg). At the end of the pulse-chase period, the patterns in the amount of 13C remaining in the above-ground tissues of the different functional types remained broadly similar to the amounts of fixed 13C (Fig. 3b), with graminoids from the controls retaining the most 13C (overall significant effect of functional type; F2,29 = 12.9; < 0.001). The proportion of 13C initially fixed by the plants remaining in above-ground tissue was similar both between treatments and between forbs and graminoids, where c. 15% of the initial fixed 13C remained (Fig. 3c). By contrast, the 13C content of bryophytes, which fixed very little 13C during the 4 h labelling period, increased by over 100% at the end of the pulse-chase period; this was reflected by an overall significant effect of functional type on %13C remaining (F2,29 = 12.2; = 0.001). By the end of the pulse-chase period, the 13C remaining in above-ground biomass, expressed as a percentage of the 13C initially fixed, accounted for 14.7% in the controls and 18.9% in the droughted cores (Table 1). The quantity of 13C recovered in roots within the control cores was 5163 μg and in the droughted cores was 2997 μg, representing 66 and 30.8% of the initial amount of 13C fixed, respectively (Table 1). The amount of 13C recovered in bulk soil accounted for 22.8% in the controls and 35.8% in the droughted cores of that which was initially fixed (Table 1).

Figure 3.

The effect of drought (hatched bars) on carbon pools in forbs, graminoids and bryophytes and the total green biomass in cores removed intact from a Nardus stricta snowbed (± SEM). (a) The amount of 13C fixed; (b) the amount of 13C remaining in each component and the total biomass at the end of the 140 h pulse chase period; and (c) the proportion of 13C remaining in each component and the total biomass expressed as a percentage of the amount initially fixed by each component or in total.

Table 1.   Recovery of 13C expressed as total quantities (μg) and proportions (%) of 13C initially fixed by vegetation and subsequently recovered in plant, soil and gaseous pools and fluxes during a 5 d pulse-chase period of intact Nardus stricta snowbed cores subjected to two simulated drought events (± SEM)
Pool or fluxQuantities recovered (μg)Recovery as a percentage of amount fixed (%)
ControlDroughtControlDrought
  1. DOC, dissolved organic carbon.

Shoots1517 ± 5061059 ± 41214.7 ± 1.118.9 ± 2.8
Roots5163 ± 19792997 ± 239766.0 ± 20.330.8 ± 13.5
Bulk soil1776 ± 5951641 ± 57322.8 ± 8.835.8 ± 14.8
DOC0.597 ± 0.150.256 ± 0.080.008 ± 0.0010.006 ± 0.002
CO2 efflux244 ± 4035 ± 5.24.5 ± 1.71.5 ± 0.6
Total8700 ± 29125732 ± 2999108 ± 30.187 ± 18.3

Losses of CO2 from the cores were quantified on two occasions during the pulse-chase period (Fig. 4a,b). ANOVA revealed an overall negative effect of drought on net CO2 exchange with significantly (F1,19 = 9.75; = 0.014) less C being lost from this pathway. By contrast, there was a significant (F1,19 = 9.59; = 0.015) treatment × time interaction on ecosystem respiration. Here, drought decreased losses of CO2 on the first sampling date but had no effect on the second. The pattern of gross photosynthesis was similar but more pronounced, with a significant (F1,19 = 6.6; = 0.033) treatment × time interaction and an overall significant effect of treatment (F1,19 = 6.11; = 0.039). On the first sampling occasion, gross photosynthesis decreased from 0.20 ± 0.002 to 0.13 ± 0.01 μg CO2-C s−1 per core. Mean gross photosynthesis tended to increase linearly with total above-ground biomass (R2 = 0.496; F1,9 = 9.9; = 0.014), and in particular the biomass of graminoids (R2 = 0.676; F1,9 = 19.8; = 0.002). When combined with measurements of the δ13C signatures, there was a significant (F1,29 = 33.0; < 0.001) reduction in 13CO2 released from ecosystem respiration in response to drought (Fig. 4c), but this was dependent on time (treatment × time interaction: F2,29 = 15.0; < 0.001). This effect was greatest on 22 July when 13CO2 losses decreased > seven times from 1.09 to 0.15 ng 13CO2-C s−1 per core. Whilst the effect of drought was still clear on 24 and 27 July, the magnitude of the losses had reduced. The loss of 13CO2 from ecosystem respiration accounted for 4.5% of the amount fixed in the controls and 1.5% in the droughted cores (Table 1).

Figure 4.

The effect of drought (hatched bars) on net CO2 exchange, ecosystem respiration and gross photosynthesis on 22 July (a) and 24 July (b), and on the efflux of recently fixed 13CO2 (c) via ecosystem respiration during the pulse-chase period (± SEM). The 4 h pulse labelling finished at 15:00 h on 21 July. Asterisks indicate significant differences within a vegetation type (< 0.05).

Losses of DOC in leachate were driven by a combination of differences in both leachate volume and C concentration between the treatments. The volume of leachate recovered during the pulse-chase period increased steadily and was significantly (F1,49 = 21.1; < 0.001) lower in control cores, but this was driven by differences in the first 3 d (Fig. 5a). The total amount of C recovered in leachate also increased throughout the pulse-chase period but was significantly lower (F1,49 = 806.8; < 0.001) in the droughted treatment (Fig. 5b), despite these cores producing more leachate in this period. This effect was maintained when the data were combined with the δ13C signatures of the DOC (Fig. 5c). The droughted cores released significantly (F1,39 = 27.6; < 0.001) less 13C in leachate. The amount of 13C recovered increased substantially in both treatments beyond the 23 July sampling date, when recoveries were two orders of magnitude greater than on the 22 July, although the droughted cores still released significantly less 13C. Despite the large treatment effect, the contribution of DOC to 13C losses for the 5 d pulse-chase period was small, representing only 0.008% of the amount of 13C fixed in the controls and 0.006% in the drought treatment (Table 1).

Figure 5.

Volume of leachate collected (a), total C recovered in leachate (b), and total 13C recovered in leachate (c) during a 140 h pulse-chase period from control (closed circles) and droughted (open circles) cores from a Nardus stricta snowbed (± SEM).

Discussion

We report the results from one of the first experiments to use 13CO2 pulse labelling to test how drought affects ecosystem C fluxes in alpine systems. We found that the plant communities subjected to drought stress fixed less 13C and allocated less 13C below ground (Fig. 6). Our data did not fully support our hypotheses; firstly, less DOC was recovered from the droughted cores even though the volumes of leachate produced during the labelling period were greater and the final drought event had occurred over 1 wk previously. This may reflect reduced water-use efficiency by the droughted plant communities that tended to have less biomass. However, we cannot rule out the possibility that the drought treatment may have affected soil physical conditions, such as the abundance of macropores that could lead to preferential flow of water through the profile. In droughted cores, recent plant assimilate (13C) also contributed less to the total DOC loss than in the controls. Secondly, losses of 13CO2 from ecosystem fluxes were significantly lower in the drought treatments. However, a striking finding was that the C fluxes in the droughted cores were associated with changes in the relative abundance of the major plant functional types, and a tendency for an overall reduction in the biomass of the plant communities. In particular, the dominant forbs (V. myrtillus, P. erecta) performed much better in the droughted cores whilst the dominant graminoid (N. stricta) performed much worse. This is surprising given the overall dominance of N. stricta in these systems and that it often roots much deeper than the most abundant forbs (Heath et al., 1938). Nevertheless, root development of N. stricta is sensitive to drought (Fiala et al., 2009) and it may be that this enables other species to be more competitive. In grassland, both biomass (Dias et al., 2010) and community composition control the amount of CO2 fixed per unit area and both these factors were probably driving the changes in 13C fixation and subsequent translocation in our experiment. Our findings demonstrate the importance of above-ground–below-ground interactions in regulating C fluxes and point to a key role of plant functional type diversity in determining how ecosystems respond to the effects of global climate change.

Figure 6.

Quantity of 13C recovered in vegetation and soil pools and ecosystem CO2 fluxes throughout the 140 h pulse-chase period. Mean losses of 13C-dissolved organic carbon (DOC) in leachate are not shown but were 0.033 μg 13C m−2 in controls and 0.014 μg 13C m−2 in the drought treatment.

Sphagnum appeared to be resilient to the drought treatment, which is surprising given the changes in canopy cover of higher plants (creating a more open habitat with probably lower humidity), greater reliance of mosses on surface water and sensitivity to natural drought events (Bragazza, 2008). Although bryophytes made only a small contribution to the amount of 13C fixed by the plant community, they retained this entire C pool. Indeed, the isotopic enrichment of moss tissue increased throughout the pulse-chase period, indicating that the plants were photosynthesizing 13CO2 released from the soil surface. This tight coupling of below-ground C allocation by vascular plants and subsequent partial fixation of autotrophic respiration by mosses may be important. In a 13CO2 pulse-labelling experiment in the high Arctic, mosses were found to allocate and retain most of their recently photosynthesized C in recalcitrant pools in contrast to co-occurring higher plants (Woodin et al., 2009). Mosses may therefore play a key role in sequestering and diverting (as highlighted in our study) otherwise relatively labile C into long-term resistant pools.

One of the clearest effects of the drought treatment in our experiment was to dramatically reduce CO2 loss from ecosystem respiration. We do not know the relative contribution of plant or soil respiration to this flux, but the reduction in the recovery of total C and 13C from leachate and reduced allocation of 13C to roots would suggest the former. This may indicate that plants remained stressed from the drought events imposed 8 d previously, even though the soil water content returned to control values and the severity of the treatment was low. Indeed, the imposition of ‘ecologically relevant’ drought treatments indicates that one of the key functions performed by alpine snowbed ecosystems, namely regulation of C turnover, is highly sensitive to stress as a result of reduced water availability. By contrast, neighbouring alpine heath plant and soil microbial communities developed on much thinner, more mineral soils have been shown to be resilient to climate warming and sustained inputs of simulated atmospheric N deposition (Papanikolaou et al., 2010).

Our results demonstrate that recent assimilate moves rapidly through several key pathways in N. stricta-dominated ecosystems. The recovery of 13C in bulk soil and in DOC suggests that labile C derived from recent photosynthate is rapidly available for mineralization by soil microorganisms. The most likely mechanisms behind these pathways are exudation of low-molecular-weight compounds (sugars, organic acids) from both roots and mycorrhizal mycelium (Jones et al., 2004). Both arbuscular mycorrhizal (AM) fungi, which colonize N. stricta roots (Read et al., 1976), and ericoid mycorrhizal fungi, which colonize the majority of root cortical cells in Vaccinium species (Pearson & Read, 1973), can be large sinks of recent assimilate (Stribley & Read, 1974; Johnson et al., 2002a,b). Further work is necessary to test recent predictions (De Deyn et al., 2008) that shifts in abundance of dominant plant functional types, here caused by the drought treatment, lead to selection of particular groups of microorganisms with differing abilities to mineralize labile C. It is also possible that the roots of N. stricta, where the extent of AM fungal colonization can be variable (Genney et al., 2001), contribute more to DOC in leachate than the fine roots of Vaccinium species that are always heavily colonized by ericoid mycorrhizal fungi.

Whilst the contribution of DOC in leachate to the overall C flux is small, this is likely to be a pool that has rapid turnover (Bengtson & Bengtsson, 2007). In addition, we only measured 13C content of leachate for 5 d after labelling finished. Because the quantity of 13C was increasing slightly, even at this final harvest, it is likely that the contribution of DOC to the overall pool would increase, while the contribution of other pathways (e.g. ecosystem respiration and shoot biomass) would decrease. Furthermore, observations of water colour from the cores indicated differences in the composition of DOC in leachates between the control and drought treatments, and analysis of their quality may provide further insight into the processes governing C release from these systems. For example, it has been shown that phenolic content of leachate can be an important regulator of DOC turnover (Freeman et al., 1996).

The change in the composition of the plant communities, with an increase in forb biomass at the expense of graminoids, is a surprising and important outcome. This finding demonstrates further the importance of consideration of above-ground–below-ground interactions (Wardle et al., 2004) in regulating ecosystem C fluxes and the need to determine how key morphological, functional and reproductive traits of alpine plants in general respond to drought stress. We emphasize that our study was conducted on material from a single snowbed; there is clearly a need to consider the response of a broad range of alpine grassland ecosystems that have adapted under different conditions to the one we focused on. Nevertheless, our study provides compelling evidence that modest drought events have measurable effects on quantities and pathways of C flux in snowbed communities. Whether these effects have consequences for C sequestration in these systems remains to be determined. Because we began to see recovery in some fluxes, notably ecosystem respiration (Fig. 4b), this suggests the direct consequences may be transient. However, this belies other potential effects of drought such as the release of C from older pools, the cumulative effects of an increased frequency of drought events and the possibility of ‘priming’ (Fontaine et al., 2007). Our finding that C turnover in snowbed communities is sensitive to drought stress indicates that increases in the frequency of short-term extreme drought events predicted for Scotland over the next century (Vidal & Wade, 2009) may have important consequences for the functioning of alpine ecosystems. To test this, we need to obtain more precise quantitative data on predicted fluctuations in soil water content in alpine plant communities that can then be directly related to experimental manipulations.

Acknowledgements

We thank Julia Fisher, Barry Thornton, Allan Sim and Leah Jackson-Blake for technical support. J.V. received a bursary from Fondation Desjardins and we also acknowledge DEFRA and NERC for financial support. A.J.B. and R.C.H. are funded by the Scottish Government Rural and Environment Research and Analysis Directorate (RERAD).

Ancillary