Can frequent precipitation moderate the impact of drought on peatmoss carbon uptake in northern peatlands?



  • Northern peatlands represent a large global carbon store that can potentially be destabilized by summer water table drawdown. Precipitation can moderate the negative impacts of water table drawdown by rewetting peatmoss (Sphagnum spp.), the ecosystem's key species. Yet, the frequency of such rewetting required for it to be effective remains unknown. We experimentally assessed the importance of precipitation frequency for Sphagnum water supply and carbon uptake during a stepwise decrease in water tables in a growth chamber.
  • CO2 exchange and the water balance were measured for intact cores of three peatmoss species (Sphagnum majus, Sphagnum balticum and Sphagnum fuscum) representative of three hydrologically distinct peatland microhabitats (hollow, lawn and hummock) and expected to differ in their water table–precipitation relationships.
  • Precipitation contributed significantly to peatmoss water supply when the water table was deep, demonstrating the importance of precipitation during drought. The ability to exploit transient resources was species-specific; S. fuscum carbon uptake increased linearly with precipitation frequency for deep water tables, whereas carbon uptake by S. balticum and S. majus was depressed at intermediate precipitation frequencies.
  • Our results highlight an important role for precipitation in carbon uptake by peatmosses. Yet, the potential to moderate the impact of drought is species-specific and dependent on the temporal distribution of precipitation.


Although northern peatlands cover only c. 3% of the earth's land surface, they accumulated an equivalent of 20% of total terrestrial soil carbon throughout the Holocene (Turunen et al., 2002; Yu, 2011; Kleinen et al., 2012), acting as a significant sink for atmospheric CO2 (Jobbágy & Jackson, 2000; Roulet et al., 2007; Nilsson et al., 2008). Whether these ecosystems will continue to function as a carbon sink in a warmer future climate is currently uncertain, calling for a mechanistic understanding of feedbacks between atmosphere and ecosystems. Climate projections for the Northern Hemisphere indicate a shift towards higher temperatures (IPCC, 2007) and less frequent, but more intense precipitation events (Allen & Ingram, 2002; O'Gorman & Schneider, 2009), which will probably lead to deeper water tables and drier surface conditions in peatlands. How these deeper water tables will affect carbon exchange and how this will interact with changes in the temporal distribution of precipitation remain to be explored (Robroek et al., 2009; Fenner & Freeman, 2011; Heijmans et al., 2013).

Changes in the temporal distribution of precipitation are of crucial importance for carbon cycling and may even have more impact on ecosystem functioning than precipitation quantity per se (Knapp et al., 2002; Heisler & Weltzin, 2006; Vervoort & van der Zee, 2008; Xu et al., 2013). This may be particularly important for northern peatlands, where a significant part of ecosystem carbon uptake is carried out by peatmosses (genus Sphagnum) (Frolking et al., 2002; Riutta et al., 2007; Kuiper et al., 2013; Street et al., 2013). Peatmosses lack stomata, water-conducting tissue and roots, making their photosynthesis and associated carbon uptake highly dependent on the water present in the living layer (top 5–10 cm) of the moss carpet (Schipperges & Rydin, 1998; Robroek et al., 2009). In the case of shallow water tables, water supply to the living moss layer is dominated by capillary rise (i.e. water transport from the water table to the peat surface) (Ketcheson & Price, 2013), and carbon uptake is not limited by water deficiency. In the case of deeper water tables, however, capillary rise strongly decreases (McCarter & Price, 2014), resulting in lower water availability and associated reduced carbon uptake by the mosses (Alm et al., 1999; Ciais et al., 2005; Aurela et al., 2007). In conditions with a deep water table, precipitation water intercepted by the living moss layer can be used for Sphagnum photosynthesis (Robroek et al., 2009; Adkinson & Humphreys, 2011). However, as water retention of Sphagnum plants is limited (Ketcheson & Price, 2013), the effect of precipitation may be strongly modified by the frequency of precipitation and the speed at which mosses can use this transient resource for photosynthesis (Campbell & Grime, 1989). Yet, these frequency-dependent effects have received limited attention in northern peatlands, with previous studies focussing on total precipitation amount, rather than on the temporal distribution of precipitation (Sonesson et al., 2002; Robroek et al., 2009; Adkinson & Humphreys, 2011; Keuper et al., 2012).

We experimentally assessed the importance of the temporal distribution of precipitation for water content, photosynthesis and carbon uptake of peatmoss during a stepwise decrease in water tables, using three species representative of northern peatlands. We hypothesized (1) that carbon uptake would decrease with deeper water tables, but that precipitation would moderate this effect. More specifically, we expected (2) carbon uptake to be larger for small but frequent events than for large infrequent events. We further expected (3) the effect of precipitation frequency on carbon uptake to be species-specific.

Material and Methods

Plant material

We selected three species representative of the three most common microhabitats (hollow, lawn and hummock) in peatlands (Table 1). Sphagnum majus (Russ.) is characteristic of (semi) continuously inundated microhabitats (hollows) and Sphagnum fuscum (Schimp.) of elevated microhabitats further from the water table (hummocks), while Sphagnum balticum (Russ.) occupies intermediate positions (lawn).

Table 1. Water table treatment levels and water contents per species
 Wet Wet field conditionsMoist Optimal field conditionsDry Drought, no capillary riseRewetted Wet field conditions
  1. The duration of each water table treatment is given in the row ‘duration’. Volumetric water contents (m3 H2O per m3 Sphagnum) per water table treatment are averaged over time and precipitation frequency treatments. Small letters indicate, per row, which water table treatments have statistically similar water contents.

Water table below moss surface (cm)
Sphagnum fuscum 25345425
Sphagnum balticum 9.4153111.1
Sphagnum majus 2.56.5152.5
Duration (d)15141717
Volumetric water content (m3 m−3)
Sphagnum fuscum 0.40a0.35b0.27c0.35b
Sphagnum balticum 0.76a0.65b0.44c0.65b
Sphagnum majus 0.95a0.93a0.73b0.94a

Moss cores were collected in September 2011 at Kulflyten, a 1-km2 ombrotrophic mire in southern Sweden (59°54′N 15°50′E; 130 m above sea level (asl); see Sjörs (1948) for a detailed site description). We collected 25 cores per species from monospecific patches (> 95% cover; < 2% vascular plant cover) spread over five different locations in the mire. The cores were obtained by gently pressing PVC cylinders (inner diameter 15 cm; height 10 cm) into the moss carpet while cutting around them with a finely serrated knife to prevent sample compression along the cylinder wall. Vascular plant shoots, if present, were removed by clipping.

After transport to the Netherlands, cores were allowed to acclimate for a 84-d period, after which treatments were randomly assigned to the cores (hereafter referred to as mesocosms). The acclimation period consisted of 44 d outside under a roof (60% shading; mean (± SD) temperature 12.2 ± 3.4°C) and 40 d in the growth chamber at experimental climate settings (mean temperature 18.5°C; see section ‘Growth chamber climate settings’) with a water table 4 cm below the moss surface.

Growth chamber experiment

We assessed the importance of precipitation frequency for water supply and carbon uptake of peatmoss in a growth chamber experiment with two factors: water table (four treatment levels) and precipitation frequency (five treatment levels) for three Sphagnum species, with five replicates for each precipitation frequency–water table treatment combination.

Water tables

Water tables were lowered in three steps through time (‘wet’, ‘moist’ and ‘dry’), simulating the seasonal drop in water tables during summer (Table 1; Fig. 1). At the end of the experiment, water tables were raised back from ‘dry’ conditions to ‘wet’ pre-drought levels (‘rewetted’), enabling assessment of the recovery of carbon uptake after summer drought (Fig. 1). The three consecutive water tables reflect natural conditions for each species during wet field conditions (‘wet’), optimal field conditions (‘moist’) and summer-drought conditions (‘dry’) (Table 1). The exact water table treatment settings were species-specific and were based on the vertical zonation of the three peatmosses along the water table under field conditions as reported in the literature (Andrus et al., 1983; Rydin, 1986; Lafleur et al., 2005; Nilsson et al., 2008).

Figure 1.

Water table treatment (upper panel) and precipitation frequency treatment (lower panel) application through time. The shaded grey areas represent the stabilization periods during which the water table and water content were allowed to equilibrate. In the lower panel, the vertical lines indicate dates of precipitation application, the length of the lines representing the precipitation amount. The black triangles indicate the timing of net ecosystem exchange (NEE) measurements, 5–30 min before rain application.

The water table treatments were imposed by changing suction at a given time. To this end, the mesocosms were placed in water retention cylinders (Fig. 2), an adaptation of the generally accepted sandbox method (Klute, 1986). Water tables were kept constant using a Mariotte bottle to correct for evaporation, and an overflow level to drain percolated precipitation (Fig. 2; for more details, see Supporting Information Methods S1). Accordingly, precipitation did not result in a rise of the groundwater level.

Figure 2.

Overview of the experimental set-up. Numbers represent components of the water retention cylinders and letters are water balance components. Mesocosms were placed in water retention cylinders on fine sand. A drainage pipe in the coarse sand communicates via a tube (5) with the groundwater reservoir (6). By lowering the position of this reservoir, the depth of the water table is increased. Evaporation from the moss surface is compensated for by the Mariotte bottle (8). Gas exchange was measured by placing the cuvette tightly over the mesocosm. Internal mixing of air was established with a fan (10). Interaction with the outer atmosphere was prevented by the rubber rim (12).

Our physical approach allows direct translation of the relationship between water content in the living moss layer and the water table to field conditions, provided that atmospheric conditions are similar. Potential differences in the suction and water content of the top moss layer may arise from differences in capillary rise between a natural field column and the moss–sand column we used. Within the suction range applied, the sand was well able to sustain water transport to the moss columns, as evidenced by the saturated conditions of the fine sand layer. (For more information and details of the physical characteristics of the sand, see Methods S1.) Furthermore, as the sand remained saturated throughout the experiment, precipitation only reduced the upward capillary flux and increased drainage in the Sphagnum mesocosms.

Water tables were maintained for a measurement period of c. 15 d (see Table 1 for details) with 6–7 d of equilibration between consecutive water tables (Fig. 1). This approach allowed the pore water pressure and associated mesocosm water content to equilibrate with the altered water table and evaporation. No substantial changes in water content were observed after these equilibration periods, indicating that equilibrium between water content and the water table was reached.

Precipitation frequency

Precipitation frequency treatments (once per 2, 4, 6 and 8 d, and no precipitation; Table 2) remained constant throughout the experimental period (Fig. 1). For each precipitation frequency treatment, the total precipitation amount and precipitation intensity were kept constant while precipitation duration was varied. This was done to avoid effects of total precipitation amount and intensity on carbon uptake. The duration and intensity of the precipitation events were representative of natural conditions and were calculated from intensity–duration–frequency diagrams (Dahlström, 2006; Methods S1). The average precipitation amount was set to compensate daily potential evaporative water losses in the growth chamber (2 mm d−1). Precipitation was applied using a peristaltic pumping system (Masterflex Console Drive 7520-47; Cole Parmer, Schiedam, the Netherlands) with 24 drip points. The pump was calibrated before precipitation application, to allow regulation of the precipitation intensity with high precision (1.18 ± 0.01 (SE) mm min−1). Natural precipitation water quality was simulated using a commonly used diluted seawater solution (Garrels & Christ, 1965).

Table 2. Precipitation characteristics for all precipitation frequency treatment levels
Precipitation frequency treatmentPrecipitation amount (mm)Precipitation duration (min)Dry spell length (d)
  1. All precipitation frequency treatments have the same precipitation intensity (1.18 mm min−1) and average daily water supply. Dry spell length is the inverse of precipitation frequency.

1/2 d4.13.472
1/4 d8.26.924
1/6 d12.310.386
1/8 d16.613.838
No precipitationNone

Growth chamber climate settings

The day : night temperature in the growth chamber was 20 : 17.5°C (10 h : 14 h) and corresponds to average 1961–1990 July conditions at the collection site (Kulflyten). The day temperature reflects the average temperature at noon, whereas the night temperature corresponds to the average temperature between 18:00 h and 06:00 h (local time; Västerås meteorological station; 59°60′N 16°46′E, Swedish Meteorological and Hydrological Institute). The relative humidity (RH) was 70% and the CO2 concentration was 400 ppm CO2. At the moss surface, the photosynthetic photon flux density (PPFD) was 246 ± 20 (SD) μmol m−2 s−1, net radiation in the daytime was 101 ± 4 (SD) W m−2 and the wind speed was ≤ 0.1 m s−1.


Water balance

To determine the importance of precipitation as a water source for moss evaporation, the water balance of all mesocosms was quantified over all water table treatments (Eqn 1).

display math(Eqn1)

For each precipitation treatment, the amount of precipitation (R) added during the measurement period was known. The capillary supply (C) and drainage (D) were obtained by weighing the capillary and drainage reservoir at the start and end of a 7-d period (Fig. 2). The change in water content in the mesocosms (ΔS) was quantified as the change in volumetric water content (m3 m−3) over the measurement period measured with moisture sensors, multiplied by the mesocosm volume. Changes in volumetric water content in the top 1–5 cm of the moss layer were assessed with EC5–H2O moisture sensors (Decagon Devices, Pullman, WA, USA) installed at 3 cm depth in one randomly selected mesocosm per species–precipitation treatment combination (n = 15). The moisture sensors measured the dielectric constant (Hillel, 2004) with a 5-min frequency. Dielectric constants were converted to volumetric water contents using species-specific calibration functions (Methods S2). As the evaporative loss from the moss surface (Ea) is the only unknown in Eqn 1, it can be derived from the water balance, together with water balance errors (ε), as Ea and ε cannot be distinguished. Negative values for evaporation and capillary fluxes (3% of data) were excluded before statistical analysis to reduce error variation: treatment patterns remained unaffected. Technical problems in the first week prevented assessment of the water balance for the wet treatment.

The water balance components in Eqn 1 were used to obtain a measure for the fraction of evaporation originating from retained precipitation (fp; Eqn 2), which we used as a measure for the precipitation dependence of Sphagnum.

display math(Eqn2)

An fp ratio of 1 represents 100% dependence on precipitation as a source of water (C = 0 mm d−1), whereas a value of 0 represents 100% dependence on capillary water supply.

Gas flux measurements

Carbon uptake was assessed by measuring the net CO2 flux per mesocosm (net ecosystem exchange (NEE)) from each species for each precipitation treatment and each water table depth. We expressed carbon uptake relative to the atmosphere, with negative values indicating net carbon sequestration by the mesocosms, and positive values net emission of carbon from the mesocosms (cf. Chapin et al., 2006). To assess the potential effects of precipitation frequency on NEE, we measured NEE in the driest conditions, that is, just before application of precipitation. For the rewetted water table treatment, we measured NEE 7–11 d after rewetting. Closed flux chambers (diameter 15 cm and height 24.3 cm, fitted with a circulating fan at 20 cm from the moss surface) were placed tightly over the mesocosms to measure CO2 fluxes using a photoacoustic multi gas analyser (type Innova 1302; Bruel and Kjær, Denmark, Nærum), connected to a multipoint sampler (CBISS MK2, 4-channel; CBISS Ltd, Tranmere, UK).

Tubes were flushed at 15 ml s−1 to enable independent sampling, and chamber measurements comprised three successive sampling points at an interval of 2 min. The accuracy of gas exchange measurements was further increased by compensating for water vapour interference. Partial pressures of CO2 and H2O entering the cuvettes were set to ambient CO2 by adding CO2-free air, and to ambient H2O by dehumidifying air to a pre-set dew point. NEE was calculated from the linear change in CO2 concentration in the chamber headspace with time.

Chlorophyll fluorescence

To directly explore the photosynthetic response of moss as a function of water content, the efficiency of photosystem II (PSII) was estimated by measuring chlorophyll fluorescence with a portable chlorophyll fluorometer (Mini-PAM; WALZ, Effeltrich, Germany). Chlorophyll fluorescence was measured at least once per water table treatment on the mesocosms with moisture sensors and under steady-state conditions (i.e. before or more than 1 d after precipitation application).

First, the dark-adapted minimal fluorescence yield (F0) was determined by illumination with far-red light. To this end, a PVC lid with 11 covered holes was placed on the samples for a 15-min period of dark adaptation. One by one, the covers of each hole were removed, after which the fibre-optic probe was immediately inserted approximately level at 0.5–2 cm above the moss surface. The maximum chlorophyll fluorescence (Fm) was then obtained by emitting an 800-ms, high-intensity saturation pulse. This procedure was repeated for the different water table treatments to obtain a broad range of water contents. To be able to compare measurements at different points in time, the lid was positioned at the same location every time. The maximum quantum yield of PSII photochemistry, Fv/Fm, can be calculated as (F– F0)/Fm and is a measure of the efficiency of PSII (Maxwell & Johnson, 2000).

To determine if water content affects the efficiency of PSII, a generalized logistic function (Eqn 3) was fitted using an adaptive nonlinear least squares algorithm (nls package, R v2.13.0; R Development Core Team, 2013) for each species.

display math(Eqn3)

Here, parameter VWCPSII50 represents the volumetric water content (VWC; m3 H2O per m3 Sphagnum) at which PSII efficiency is 50% and switches from active to inactive, while the parameter β represents the steepness of this switch.

Chlorophyll a + b content

To determine if treatment effects on carbon uptake could be attributed to damaged chloroplasts and decreased photosynthetic capacity, the chlorophyll content was analysed destructively at the end of the experiment. From each mesocosm, the capitulums (top 1 cm) of five random Sphagnum shoots were collected, snap-frozen in liquid nitrogen, and stored at −70°C. Next, the samples were freeze-dried and ground, after which chlorophyll a and b were extracted with a 96% ethanol solvent and their contents determined spectrophotometrically using specific absorption coefficients and equations as described by Lichtenthaler (1987).

Data analysis

All data were tested for normality (Shapiro–Wilk) and equality of variances (Levene's test). As water table treatment levels were species-specific, treatment effects were tested for each species separately. Treatment effects on precipitation dependence (fp) and NEE were tested with full factorial linear mixed models with precipitation frequency as a fixed factor, water table treatment as a within-subjects factor and mesocosm as a random effect using spss (v19.0.0.1; SPSS/IBM Inc., Somers, NY, USA). Likelihood ratio tests were performed to select the most parsimonious covariance structure from a set of covariance structures that account for correlation and heterogeneous variances. If significant interactions (P < 0.05) between precipitation frequency and water table were present, the effect of precipitation frequency on fp and NEE was tested with separate one-way ANOVAs for each water table treatment. Multiple comparisons were carried out per species and water table treatment to determine which precipitation frequency treatments differed significantly from each other. P-values were corrected for multiple comparisons using the Benjamini–Hochberg method (Waite & Campbell, 2006). In the analysis of both fp and NEE, the first-order autoregressive covariance structure was the most parsimonious for all species and was accordingly adopted in all linear mixed models.

Recovery of NEE after drought was quantified by comparing NEE in the wet and rewetted water table treatments (Table 1). To investigate whether more frequent precipitation enhanced recovery of NEE, repeated-measures ANOVA was performed per species. Precipitation was included as a between-subjects factor, water table treatment (wet versus rewetted) as a within-subjects factor and NEE as the dependent variable. Separate paired sample t-tests were performed per precipitation frequency to check if the effect of water table on NEE interacted with precipitation frequency.

To also check whether water availability affected NEE and to determine how this response was reflected in PSII efficiency, regression analyses were performed with NEE or PSII as the dependent variable and water content as the independent variable. Sphagnum carbon uptake is known to be decreased under reduced water availability, but also in the near-saturation range as a result of the reduced diffusivity of CO2 in water (Williams & Flanagan, 1996; Schipperges & Rydin, 1998). Hence, the response of carbon uptake to water availability was expected to be unimodal but not necessarily symmetrical. Accordingly, first- to third-order polynomials were fitted and the most parsimonious variant was selected with likelihood ratio tests. Data are available from the repository (Nijp et al., 2014).

Differences in mean total capitulum chlorophyll content between precipitation frequency treatments at the end of the experiment were determined with a one-way ANOVA for each species.


Precipitation dependence

The water table treatments successfully imposed differences in the water contents of the living moss layer representative of field conditions (Table 1) for all species. Precipitation tended to increase water contents in the top layer, particularly when the water table was deep (data not shown). This effect could not be further quantified, however, because of the limited number of moisture sensors (one per precipitation treatment per species). Precipitation dependence (fp) increased significantly with increased water table depth, irrespective of species (Fig. 3, Table 3). Precipitation dependence was significantly higher in dry conditions (deep water table treatment; Table 1) than in moist conditions (intermediate water table treatment) for all species (Benjamini–Hochberg (BH) corrected multiple comparisons, P < 0.05). Overall, rainwater dependence for all species increased by 37% between moist and dry conditions. In moist conditions, 8–34% of evaporation originated from precipitation, whereas under dry conditions this shifted to 53–67%, with S. majus showing the lowest precipitation dependence and S. balticum the highest precipitation dependence. For S. fuscum, the imposed water tables resulted in intermediate values (23% under moist conditions and 53% under dry conditions). Precipitation frequency had a limited influence on precipitation dependence. The only significant effect was found for S. balticum, where precipitation dependence was higher at intermediate precipitation frequencies (BH multiple comparisons, P < 0.05).

Table 3. Significance of precipitation frequency (Freq) and water table (WT) effects on precipitation dependence (fp) per species
Effect Sphagnum fuscum Sphagnum balticum Sphagnum majus
F df P F df P F df P
  1. The fp indicates whether groundwater (fp < 0.5) or precipitation (fp > 0.5) dominates water supply. WT is treated as a within-subject effect. Bold values indicate significant effects (P < 0.05). See Fig 3 for interaction effects and significant subgroups (Benjamini–Hochberg corrected multiple comparisons) of precipitation frequency per water table treatment level.

Intercept3171, 13.5 < 0.001 95.51, 15.9 < 0.001 3091, 12.8 < 0.001
Freq2.723, 13.40.08633.63, 15.5 0.037 2.133, 12.60.148
WT62.71, 16.2 < 0.001 24.21, 15.6 < 0.001 5.831, 13.9 0.030
Freq × WT1.263, 16.60.3200.403, 15.00.7520.743, 13.70.545
Figure 3.

Interactive effect of water table and precipitation frequency treatments (greyscale) on precipitation dependence (fraction of evaporation from retained precipitation water (fp)) of three Sphagnum species. fp values indicate whether groundwater (fp < 0.5) or precipitation (fp > 0.5) dominates water supply. Letters represent homogeneous subgroups of precipitation frequency treatments per water table (Benjamini–Hochberg corrected multiple comparisons). Error bars, ± 1 SE. See Table 3 for significance of main and interaction effects. ns, not significant.

Carbon uptake

Net ecosystem exchange (NEE) was significantly affected by both the water table and precipitation frequency for all species (Table 4, Fig. 4). In general, precipitation compensated the adverse effects of deep water tables (dry conditions) on carbon uptake and interacted with precipitation frequency (Fig. 4). For all species, frequent precipitation seemed to moderate the impact of drought on carbon uptake. However, the exact effect of precipitation frequency on carbon uptake was species-specific and depended on the water table. For S. fuscum, the precipitation frequency did not affect carbon uptake under wet conditions (one-way ANOVA; F4,19 = 0.80; P = 0.54; BH multiple comparisons, P > 0.05), but significantly increased carbon uptake under dry conditions (one-way ANOVA; F4,20 = 4.41; P = 0.01; BH multiple comparisons, P < 0.05). Similar to S. fuscum, the carbon uptake of S. balticum and S. majus remained unaffected by precipitation frequency in wet conditions (one-way ANOVA; F4,18 ≤ 0.709; P ≥ 0.596). In dry conditions, however, the carbon uptake of these species responded nonlinearly to precipitation frequency. For S. balticum, for example, carbon uptake was significantly lower in the 1/4 d than the 1/2 d precipitation frequency treatment, indicating that carbon uptake was depressed at this intermediate precipitation frequency (Fig. 4; BH multiple comparisons, P < 0.05).

Table 4. Effects of precipitation frequency (Freq) and water table (WT) on net ecosystem exchange (NEE)
Effect Sphagnum fuscum Sphagnum balticum Sphagnum majus
F df P F df P F df P
  1. Bold values indicate significant effects (P < 0.05) and WT is treated as a within-subject effect. See Fig. 4 for interaction effects and significant subgroups (Benjamini–Hochberg corrected multiple comparison) of precipitation frequency per water table treatment level.

Intercept2331, 20.2 < 0.001 94.61, 20.5 < 0.001 5801, 19.5 < 0.001
Freq3.194, 20.1 0.035 2.694, 20.40.0602.634, 19.50.066
WT22.02, 38.9 < 0.001 1662, 32.8 < 0.001 1572, 36.7 < 0.001
Freq × WT2.208, 38.8 0.049 2.658, 32.7 0.023 3.128, 36.7 0.009
Figure 4.

Mean net ecosystem exchange (NEE) per treatment per precipitation frequency for three Sphagnum species. NEE values are expressed relative to the atmosphere, with negative values indicating net carbon sequestration by the mesocosms, and positive values net emission of carbon from the mesocosms (cf. Chapin et al., 2006). Error bars represent ± 1 SE and letters indicate significant differences between precipitation frequency treatments within each water table treatment (Benjamini–Hochberg corrected multiple comparisons). See Table 4 for statistics for main and interaction effects. ns, not significant.

Water, carbon uptake and photosystem efficiency

To explore the existence of critical moisture thresholds, we expressed carbon uptake and efficiency of PSII as a function of the VWC of the living moss layer, combining all water table and precipitation treatments. In a stepwise polynomial regression with VWC as the explanatory variable, 28% of the variation in carbon uptake for S. balticum could be explained by VWC. Despite the limited explained variation, the fitted parameters were significant (P < 0.05) and S. balticum switched from carbon uptake to carbon emission at a mean (± SE) water content of 0.48 m3 m−3 ± 0.04 (polynomial regression; see Fig. 5a). The water content at which the switch between carbon uptake and emission occurred corresponded to the water content at which PSII efficiency reduced sharply (mean (± SE) 0.49 m3 m−3 ± 0.02; regression of Eqn 3; P < 0.001; Fig. 5b), showing that photosynthesis dominated the NEE response of the mesocosms. The water content threshold at which photosynthesis practically stopped corresponded to a water table of c. −30 cm (Table 1). The analysis suggests that the range of water contents imposed by the treatments was not large enough to reliably derive PSII efficiency response curves for S. fuscum and S. majus. Nonetheless, these species also showed comparable trends in PSII efficiency as a function of water content. With decreasing water content, first the PSII efficiency of S. majus approached the inactive state, followed by that of S. balticum and then S. fuscum.

Figure 5.

(a) Relationship between net ecosystem exchange (NEE) and volumetric water content (VWC) for Sphagnum balticum. Negative NEE represents Sphagnum carbon uptake. The solid line represents a third-order polynomial without intercept (F3,52 = 23.9; R2adj = 0.28; P < 0.001; NEE = 7.6VWC − 23.8VWC2 + 16.5 VWC3), the dashed lines represent 95% confidence intervals and the vertical dotted line represents the water content at which carbon uptake switches to emission. (b) Photosystem II efficiency (PSII) as a function of VWC for S. balticum. A generalized logistic function (Eqn 3;math formula) was fitted (solid line). The FMIN parameter was not significant and therefore excluded from the model, but all other parameters were highly significant (P < 0.001; R2adj = 0.55). The dashed and dotted lines represent the 95% confidence bounds and the water content at which the steepest decline in PSII efficiency takes place, respectively.

Recovery after drought

Eleven days after the water table was raised to restore wet conditions (the rewetted water table treatment), carbon uptake (almost) fully recovered to values obtained initially under wet conditions for S. fuscum (P = 0.081; paired samples t-test). For S. majus and S. balticum, carbon uptake was still significantly lower, suggesting lag effects of drought on carbon uptake (Fig. 6, Table 5).

Table 5. Effects of rewetting after drought and precipitation frequency (Freq) on recovery of net carbon exchange (NEE)
Effect Sphagnum fuscum Sphagnum balticum Sphagnum majus
F df P F df P F df P
  1. Recovery was quantified by comparing the wet water table (WT) treatment with the rewetted water table treatment for each precipitation frequency in a repeated measures ANOVA. Bold values indicate significant effects at the 0.05 level and WT was included as a within-subjects factor.

Intercept1071, 19 < 0.001 1621, 18 < 0.001 3211, 20 < 0.001
Freq1.354, 190.2891.304, 180.3082.984, 20 0.044
WT3.401, 190.08161.71, 18 < 0.001 30.21, 20 < 0.001
Freq × WT0.654, 190.6314.324, 18 0.013 1.284, 200.310
Figure 6.

Net ecosystem exchange (NEE) before (wet water table treatment; dark grey bars) and after (rewetted water table treatment; light grey bars) rewetting per Sphagnum species to identify recovery of carbon uptake after drought. Negative NNE values denote Sphagnum carbon uptake. Error bars represent ± 1 SE and symbols the significance of the difference in carbon uptake between the two water table treatments per precipitation frequency as determined with paired-sample t-tests (**, P < 0.01; *, P < 0.05; (*), P < 0.10; ns, not significant). See Table 5 for the significance of main and interaction effects.

Recovery of carbon uptake of S. balticum increased with precipitation frequency, illustrating the importance of frequent precipitation for longer term carbon uptake of this species. For S. fuscum and S. majus, however, recovery remained unaffected by precipitation frequency (Table 5). For all species, recovery was unrelated to NEE in the dry treatment (R2 ≤ 0.07; P > 0.19) and was not significantly related to chlorophyll content. Average chlorophyll contents for each precipitation frequency (mean ± SE: S. fuscum, 1.26 ± 0.09; S. balticum, 0.81 ± 0.06; S. majus, 1.63 ± 0.13 mg chlorophyll a + b per gram dried Sphagnum) were well within the range of field conditions for Sphagnum (Marschall & Proctor, 2004; Granath et al., 2009).


Precipitation frequency is important when the water table is deep

We have shown that precipitation becomes an important source of water for Sphagnum plants when the water table is deep, supplementing capillary water supply. In summer-drought conditions, we found that the relative importance of precipitation as a water source for peatmoss was on average 37% higher than for optimal field conditions (Fig. 3, Table 3), irrespective of precipitation frequency. Despite its negligible effect on Sphagnum water supply, however, precipitation frequency did affect carbon uptake when the water table was deep, with frequent precipitation (once per 2 d) partly offsetting the negative effects of deep water tables for all species considered in this study. The imposed drought conditions in this study are characteristic of average July conditions in southern Sweden and hence represent a relatively mild drought. Consequently, our results provide a conservative estimate of the importance of precipitation for moderating the impact of drought on carbon uptake of peatmoss in northern peatlands.

In ecosystem models, the water table is often implemented as the only representation of water availability in the living moss layer and (in)directly linked to Sphagnum carbon uptake (Yurova et al., 2007; Heijmans et al., 2008; Turetsky et al., 2012). Our results suggest that the predictive power of such models is reduced when deep water tables prevail, as precipitation becomes the dominant source of water and capillary supply by groundwater is only of secondary importance.

Response of carbon uptake to precipitation frequency is species-specific

The relationship between precipitation frequency and carbon uptake differed between species. During drought, carbon uptake of S. fuscum increased linearly with increasing precipitation frequency (Fig. 4), whereas S. balticum and S. majus responded nonlinearly, showing decreased carbon uptake, or even release, at intermediate precipitation frequency. The mechanisms underlying these species-specific responses are unclear and may be related (1) to short-term heterotrophic respiration responses after rewetting or (2) to species-specific strategies to deal with transient water supply. Short-term heterotrophic respiration responses after rewetting, also known as resaturation respiration, are generally restricted to 2–24 h after a rewetting event (Smith & Molesworth, 1973; Lee et al., 2004; Unger et al., 2010). As there was at least 48 h between precipitation events and carbon uptake (NEE) measurements, the contribution of heterotrophic respiration response seems limited.

An alternative explanation is that the species-specific response of carbon uptake to precipitation frequency is related to differences in strategies to deal with transient water supply and the time needed to reactivate photosynthesis after rewetting. Rewetting for an insufficient period of time could lead to incomplete recovery, no time for significant growth to maintain a positive carbon balance and flushing away of valuable metabolic compounds released after membrane rupture (Dilks & Proctor, 1976; Gupta, 1977; Gerdol et al., 1996; Proctor et al., 2007). As a consequence, infrequent precipitation during summer droughts could potentially intensify the negative effect of drought on carbon uptake of species with a slow response to rewetting, such as lawn and hollow species. This counterintuitive response is in line with work by Proctor & Tuba (2002). These authors linked species water use strategy to habitat and suggested that species that respond quickly to rewetting generally dominate exposed habitats with erratic water supply at fine time-scales (‘low-inertia’ species), while species that respond slowly to rewetting occupy habitats with a more predictable water availability on coarser time-scales (‘high-inertia’ species). Applying this concept to northern peatlands, we find that the species with a quick response time (S. fuscum) indeed occupies the most exposed (hummock) habitat, whereas species responding slowly to rewetting (S. balticum and S. majus) occupy less exposed lawn and hollow habitats (Fig. 4). A higher drought tolerance for hummock peatmoss species than lawn or hollow species has also been observed by Hájek & Beckett (2008). If we consider recovery of carbon uptake at longer time-scales, we again see a contrasting behaviour between hummock (S. fuscum) and lawn/hollow (S. balticum and S. majus) species. Eleven days after raising the water table to its level under pre-drought wet conditions (Table 1), carbon uptake had almost fully recovered for S. fuscum only. This suggests that species that are able to quickly switch from a photosynthetically inactive to an active state after rewetting (i.e. hummock species) may have a competitive advantage over species that take longer to recover (lawn and hollow species) when precipitation becomes less frequent but more intense.

Although it seems reasonable to assume that the reduced carbon uptake after drought observed for hollow and lawn species was a result of long-term, desiccation-induced damage to the photosynthetic apparatus, chlorophyll contents at the end of the experiment were within the range observed for optimal field conditions (Marschall & Proctor, 2004; Granath et al., 2009), indicating that either denaturation of chlorophyll did not occur or that chlorophyll was resynthesized in the post-drought period (11 d). Precipitation frequency did not affect recovery of carbon uptake, except for S. balticum, where recovery of carbon uptake increased with precipitation frequency.

The observed effects of water table and precipitation frequency on carbon uptake were based on CO2 flux measurements at one point in time, just before precipitation application. Hence, the possibility cannot be excluded that time-integrated, frequent measurements of carbon uptake throughout a drying–wetting cycle would yield different patterns. Nonetheless, a limited number of carbon exchange measurements (data not shown) with higher frequency throughout a few rewetting cycles suggested that a precipitation event was generally followed by a small respiration burst, followed by a stabilization of carbon exchange within 6 h after rewetting. This quick stabilization suggests that the patterns we found at longer time-scales will probably remain unaffected by more frequent measuring. However, we encourage other workers to test how carbon uptake responds to precipitation frequency at an even finer temporal resolution, and to upscale such findings over longer time-scales and larger spatial scales.

Relating photosynthesis to water availability

For one species (S. balticum) we were able to identify a critical moisture threshold for photosynthetic efficiency that coincided with the point at which net ecosystem exchange of S. balticum shifted from carbon uptake to emission (Fig. 5b). Although photosynthetic efficiency around this break-even point varied, this illustrates that moss photosynthesis and carbon exchange of the living peatmoss layer of peatlands are closely connected. Identifying such species-specific moisture thresholds is of crucial importance in predicting the impact of climate change on carbon uptake in northern peatlands (Strack et al., 2009; le Roux et al., 2013).

In this study, we showed that precipitation can moderate the impact of drought on peatmoss carbon uptake, but that the temporal distribution of precipitation, species identity and water table depth modify this response. These results imply that processes emerging at small spatiotemporal scales at the peat–atmosphere interface are crucial in understanding how carbon uptake of peatmosses and, ultimately, peatlands will respond to altered precipitation regimes.


This study was made possible by the Schure-Beijerinck-Popping fund (KNAW) and the Dutch Foundation for the Conservation of Irish Bogs. B.J.M.R. was supported through the Division for Earth and Life Sciences (ALW) with financial aid from the Netherlands Organization for Scientific Research (NWO; Research Innovation Scheme 863.10.014). We are indebted to Fia Bengtsson and Håkan Rydin for sharing their knowledge of Sphagnum species identification. We thank Bingxi Li, Dirk-Jan Pasma, Frans Möller, Gilian van Duijvendijk, Hennie Gertsen, Harm Gooren, Huib van Veen, Jan van Walsem, Jasper Wubs, Stijn Schreven, Stijn van Gils and Suzanne Okken for their help during different stages of the experiment.