Introduction

Concerns about the effects of elevated atmospheric carbon dioxide (eCO2) on the global carbon budget and the world’s climate have stimulated research of diverse terrestrial habitats (Ainsworth and Long 2005; Ainsworth and Rogers 2007; Leakey and others 2009; Calfapietra and others 2010). However, little is known about the effects of eCO2 on wetlands. Wetlands include a wide variety of semiaquatic ecosystems that have large variability in hydrological and ecological patterns and processes (Mitsch and Gosselink 2000). Thus, different types of wetlands may respond differently to eCO2. Although wetlands comprise only about 5–8% of terrestrial landscapes, they are considered net carbon sinks, in that they accumulate about 830 Tg C/year (12% of the estimated 7.0 Pg/year from fossil fuel combustion; Mitsch and others 2013). Wetlands may therefore play an important role in the mitigation of climate warming, although the complex interactions between hydrological and ecological factors on wetland responses to eCO2 are largely unknown. Therefore, further research is required to explain how the world’s diverse wetlands may influence the mitigation of eCO2 and global carbon balance over the next 100 years.

Previous studies have reported that several of the ecological effects of eCO2 in terrestrial plants also occur in submerged and floating macrophytes in wetlands, such as alterations of photosynthesis, plant growth, net primary productivity, acclimation, plant chemical composition, and decomposition of plant litter (Idso and Kimball 2001; Drake and others 1997; Idso 1997; Lee and others 1998; Luo and Mooney 1999; Norby and others 1999; Oren and others 2001; Marissink and others 2002; Shaw and others 2002; Drake and Rasse 2003; Hoosbeek and others 2004; Andersen and others 2006; Andersen and Andersen 2006; Yan and others 2006; Millard and others 2007). Experiments with emergent macrophytes (Scirpus olneyi, Spartina patens, Equisetum fluviatile, and Phragmites australis) showed that the photosynthetic rate and above-ground biomass increased with eCO2 and that this effect was sustained in the long term and was even greater when eCO2 was combined with increased temperature or nitrogen fertilization (Matamala and Drake 1999; Ojala and others 2002; Rasse and others 2005; Mozdzer and Megonigal 2012; Eller and others 2014). However, these experiments were carried out in growth chambers and greenhouses. Although this approach is an important first step, alterations in the growth chamber environment may artificially enhance plant growth (Kimball and others 1997; De Graaff and others 2006).

Other approaches, such as free-air CO2 enrichment (FACE), which was developed to study the effect of whole ecosystems to eCO2 (Calfapietra and others 2010) without growth chamber effects (Hendrey and Kimball 1994), were used to study bogs and rice ecosystems (Hoosbeek and others 2001; Miglietta and others 2001; Okada and others 2001; Guo and others 2012). Although freshwater (inland) marshes account for about 20% (≈ 100 million ha) of all wetlands in the world (Mitsch and Gosselink 2000), FACE has not yet been used in marshes. Marshes are wetlands with high ecological complexity associated with soil–water–plant–atmosphere processes that are strongly influenced by hydrological conditions (Reddy and DeLaune 2008). The carbon balance of wetlands under eCO2 must be assessed through integrative and realistic in situ experiments to improve our knowledge of the global carbon cycle.

Some of the early eCO2 experiments reported imbalances between the increased carbon uptake from increased photosynthesis and ecosystem carbon stocks (Rogers and others 1994). In particular, eCO2 stimulates plant growth, but also requires greater below-ground resources (water and mineral nutrients); when these demands are not met by increased resource availability or efficiency of resource use, or if growth potential is constrained, plants increase carbon loss through rapid root respiration or carbon exudation (Hungate and others 1997). Short-term experiments in wetlands have examined the emergent macrophyte E. fluviatile growing at eCO2 and demonstrated that, despite the increase in net photosynthesis, shoot and root biomass did not increase. In these studies, most of the carbon was released to the soil due to a higher turnover rate or enhanced root exudation (Ojala and others 2002). The increased carbon exudation by macrophyte roots is related to low nutrient availability (Wu and others 2012). However, the mechanism of this process under eCO2 is unclear, because nutrient depletion is unlikely to occur in the short term (Temperton and others 2003; Langley and Megonigal 2010). Although an estimated 5–21% of all photosynthetically fixed carbon is transferred to the rhizosphere through root exudates (Walker and others 2003), this increased input of carbon to the soil in response to eCO2 is considered negligible in most ecosystems. In fact, this process is only important in tropical ecosystems (Körner and Arnone 1992; Grace and others 2014). In situ experiments that consider all compartments of the wetland soil–plant environment are needed to better understand their interactions under eCO2.

The instantaneous transpiration efficiency (ITE, ratio of net photosynthesis to transpiration) indicates the efficiency of resource use by plants and increases with rising CO2 (Drake and others 1997; Barton and others 2012) due to decreased stomatal conductance (Xu and others 2016). Some studies of the combined effect of eCO2 and water availability in forests and grasslands reported that the decrease in stomatal conductance is site specific, in that it depends on environmental factors and plant species (Marissink and others 2002; Hymus and others 2003; Rasse and others 2005; Warren and others 2011; Xu and others 2016). The greatest decreases in stomatal conductance (30–40%) were reported for C3 grasses (Ainsworth and Rogers 2007).

The patterns observed in forest and grassland ecosystems are not consistently observed in wetlands, due to the pronounced effects of hydrological variability on ecological patterns and processes in wetlands (Sánchez-Carrillo and Alvarez-Cobelas 2001). Water-related stresses (excessive flooding or drought) significantly affect plant function (Sánchez-Carrillo and others 2004; Pezeshki and DeLaune 2012) and microbial responses (Reddy and DeLaune 2008). Therefore, it is likely that the seasonal variability of water level in wetlands (hydroperiod) and alterations of hydrological patterns due to global changes lead to complex ecosystem responses to eCO2. However, little is known about the combined influence of hydrology and eCO2 on wetlands under in situ conditions. For example, the rise of atmospheric CO2 may act synergistically or antagonistically with other stressors (Rasse and others 2005).

Common reed, P. australis, has a worldwide distribution and fulfills multiple roles and wetland services (Kiviat 2013), and reed beds are highly valued and protected in Europe. However, P. australis is considered an invasive species in North America because it threatens biodiversity and is therefore aggressively managed in natural areas (Tewksbury and others 2002). It is important to understand how common reed responds to rising atmospheric CO2 to predict the future evolution of wetlands under global climate change, to understand feedback mechanisms in global carbon budgets, and to improve the management of wetlands. Atmospheric CO2 concentration can affect plant and ecosystem responses, and interactions between plants and the complex interplay diverse environmental variables can modulate CO2 levels. We used a FACE facility to test the effects of eCO2 [+ 184 µmol mol−1] on P. australis in a marshland in Spain over two growing seasons. Specifically, we aimed to test the effect of this additional carbon uptake on carbon pools and turnover times and whether inter-annual differences occur due to plant aging and seasonal variability of biogeochemical, hydrological, and ecological factors. Therefore, we hypothesized that other variables such as hydrological variability and plant age modulate the wetland response under eCO2, resulting in a more ambiguous soil–plant response than was previously assumed.

Materials and Methods

FACE Site Description

The FACE experimental site (SAWFACE) is located in Las Tablas de Daimiel National Park, in central Spain (TDNP; 39°08′N, 3°43′W). TDNP is a semiarid floodplain wetland (maximum inundated area: 16 km2; average water depth: 0.90 m) with more than 90% coverage by P. australis (Cav.) Trin. ex Steudel (common reed) and less than 5% by Cladium mariscus (L.) Pohl (cut sedge) (Cirujano and others 2010). The average maximum height of reeds is approximately 1.6 m in the inundated soils and approximately 2 m in dry zones (Ortíz-Llorente 2013). From 2004 to 2009, this wetland experienced one of the severest droughts of the last 50 years, and this was followed by an extremely wet period (2010–2014), in which the wetland had the highest yearly average water level in the last 30 years. Sánchez-Carrillo and Angeler (2010) provided detailed information about this wetland.

Sánchez-Carrillo and others (2015) provided detailed descriptions and the performance of the FACE facility. Briefly, the FACE facility consists of 6 octagonal CO2 enrichment rings (eCO2 plots, 7 m2) which receive a mean CO2 of 582 µmol mol−1 and 6 control plots which receive present-day CO2 levels (399 µmol mol−1). Three of the controls are inside the facility (Control) and 3 are outside (Ext-Control), but close to the FACE area. The observed differences in macrophyte growth variables between the Control and Ext-Control were not significant, so we included them together into a single control group. The FACE experiment was installed in an area covered by P. australis. Due to extensive flooding of the wetland from December 2009 to July 2011, the reeds did not sprout spontaneously in the experimental FACE plots during the two growing seasons (2010 and 2011) preceding this experiment. To ensure that the plant material used in the experiments was uniform and that the observations were not biased due to use of plants at different developmental stages (for example, different levels of below-ground biomass), we cultivated reeds in pots from rhizomes that were collected at the beginning of March 2012 from seasonally inundated littoral areas in TDNP. After 1.5 months of acclimation, these rhizomes were planted in the FACE plots in April 2012. All collected plant materials were carefully selected and had similar morphology (length: ca. 15 cm, diameter: ca. 5 cm, rhizomes with buds from a maximum of 4 clones). Each rhizome was planted at a depth of 10 cm in 15-l polyethylene pots containing 1 kg of peat that was collected from TDNP, close to the FACE facility. For transplantation, the plastic pots were removed to expose the root balls, and the reeds were then planted without removing soil from the balls. A total of 15 germinated reed rhizomes were planted in each plot, which contained about 30–50 fully developed shoots. Before starting the initial fertilization cycle, the reeds were allowed to acclimate to the new environment. The fertilization period occurred from May to October during the 2012 and 2013 growing seasons, and CO2 additions were made during day and night.

Experimental Procedures and Statistical Analyses

Sampling was performed every 2–3 weeks from May to July, and every 4 weeks from August to October, depending on the variable being measured (see below). The FACE facility was operating 86–94% of the time during the entire experimental period, with pauses of less than 3 full days (due to equipment maintenance or malfunction).

Macrophyte growth was determined by measuring changes in the number of shoots, plant height, leaf area index (LAI), and above-ground biomass (Abiomass) during the growing period. The total number of shoots was counted by dividing plots into different portions, using ropes. As demonstrated by previous studies of P. australis in this wetland, the mean reed height in a 1-m2 plot can be accurately assessed by single measurements of maximum and minimum plant heights using a tape measure (Ortíz-Llorente 2013). Because we examined temporal changes in plant growth during the entire fertilization period (that is, 2 growing seasons), we used indirect measurements of LAI and Abiomass that did not require destructive harvesting. LAI was measured in the field (LAIcepto) using an AccuPAR LP-80 Ceptometer (Decagon Devices Inc.), in which the leaf distribution parameter x (distribution of leaf angles within the canopy) was fixed to 1.2 (Decagon 2001). These data allowed calculation of LAItrue and Abiomass using empirical models.

These empirical models were developed during the 2011 plant growing season, to consider the effects of plant phenology, and were based on 42 randomly selected 1 m2 plots of P. australis (6 plots per date: March 3, March 23, April 20, May 26, July 13, August 18, and September 27). First, LAIcepto was measured, and then plants were harvested to measure LAItrue and Abiomass in the laboratory. Triplicate LAIcepto measurements in each plot (3 records per plot on each) were taken by positioning the ceptometer N–S, E–W, NW–SE, NE–SW at 10 cm from the wetland soil. LAItrue was determined through repeated measurements of single leaves using a digital scanner and image software (Adobe Photoshop CS5, Adobe Systems Inc, for pixel color transformations, and ArcGis 10, Esri Inc., for pixel counting and area estimation). Abiomass was finally estimated after drying all plant material (leaves and stems) at 70°C for 48 h. The measured LAItrue and Abiomass had significant correlations with Abiomass (Abiomass = 108.52 × LAI2.18 (n = 42, R 2 = 0.99, SE = 12.50, p < 0.001)). Table 1 shows the regression models used to estimate LAItrue and Abiomass from LAIcepto. Abiomass was log-transformed to enable linear cross-validation, and both models were assessed using leave-one-out cross-validation (Figure 1). The Nash–Sutcliffe model efficiency coefficient (Nash and Sutcliffe 1970) for the regression models indicated both models had high predictive power (Table 1). The variability of LAItrue and Abiomass estimates (replicate variability) was included within standard errors of the empirical models.

Table 1 Regression Models Used to Estimate LAItrue [–] and Abiomass [log(g DW m−2)].
Full size table
Figure 1
figure 1

Regression models used to estimate LAItrue and Abiomass from LAIcepto. A, B Model fit and prediction intervals (PI), C, D Observed and predicted values with a comparison of fitted and cross-validated values, and E, F the residuals.

Full size image

Leaf-level photosynthesis, transpiration, and stomatal conductance were measured using a portable infrared gas analyzer (IRGA) model 225 MK3 (ADC BioScientific). On each sampling date, 5 reed stems were selected randomly in each plot, and IRGA measurements were taken randomly in the middle zone of 5 individual reed leaves, as described by Field and others (1998). All measurements were taken in the middle of the day (11–13 h), when there is maximal photosynthesis in P. australis in TDNP (Ortíz-Llorente 2013). Because transpiration and photosynthesis displayed strong seasonality, cumulative sums of transpiration and photosynthesis were calculated yearly (growing seasons) rather than annual averages.

The ITE (the ratio of net photosynthesis to transpiration) was used to test the efficiency of resource use by reed plants and its relationship with atmospheric CO2 concentration. Ten leaf samples were taken from in each plot and maintained at below 4°C during transport to the laboratory. In the laboratory, 2 leaf disks (10 mm diameter) per leaf sample were extracted to determine chlorophyll a + b [Chl (a + b)] levels after extraction with methanol. Extinction coefficients were determined using the equation of Porra and others (1989) and measurements of absorbance at 652, 665.2, and 750 nm. Foliar carbon and nitrogen were determined in five leaf samples from each experimental plot on each sampling date using an Elemental Analyzer Series II 2400 (PerkinElmer). Leaf phosphorus was measured using the molybdate/ascorbic acid method (John 1970) after sulfuric acid/ammonium persulfate wet digestion, with measurement of absorbance at 660 nm. Soil organic carbon (determined by subtracting the inorganic carbon content [measured with a Bernard calcimeter] from the total carbon content), organic nitrogen, and total phosphorus were determined in triplicate soil samples (depth: 0–20 cm, to assure the presence of any detrital material recently incorporated into the soil) from each plot before and after both fertilization cycles.

The Shapiro–Wilk’s W test was used to determine the normality of distributions and Levene’s test to determine the equality of variance and homoscedasticity. Data were log- or square root-transformed if the distribution was non-normal. Mostly, block level variability (that is, variability among plots within the same treatment group) was not statistically significant (two-sample t test: p > 0.1), and results are shown as aggregate means unless otherwise noted. A mixed-model repeated measures analysis of variance (ANOVA) was used to test for significant differences between the main effects of eCO2, time, and their interaction with plant variables, by use of a randomized complete-block design to account for spatial effects. Treatment, time, and treatment × time interaction were the fixed factors, and the block was a random factor. The Bonferroni correction was used for post hoc analyses, except for the LAItrue and Abiomass ANOVA models. Although our empirical models (which used LAItrue and Abiomass estimations) were robust, we used the Holm–Bonferroni sequential correction (Holm 1979) to decrease Type I errors, which might occur due to variability within the empirical models (that is, LAItrue and Abiomass) and variability among replicates. Based on the known phenology of P. australis, June and October records were compared to test for the cumulative effects of CO2 fertilization on plant growth. LAI and Abiomass variables were analyzed together because one is derived from the other. The Bonferroni correction was used for post hoc analyses in all other variables (stomatal conductance, leaf-level transpiration, photosynthesis, ITE, Chl[a + b], and leaf and soil carbon, nitrogen, and phosphorus). Pearson product-moment correlation analyses were used to assess the relationships between ITE/transpiration responses and Abiomass changes. To simplify data presentation, means and standard deviations reported in some tables and graphs were computed by pooling data from the six eCO2 and control replicates for a given date. Data analysis was performed using Statistica 7.0 (Statsoft Inc.) and Minitab 14 (Minitab Inc.).

Results

Number of Shoots, Plant Height, LAI, and Above-Ground Biomass

Figure 2A shows the changes in the mean water level in the FACE facility during the experimental period. During both fertilization cycles, the enriched and control plots had no significant differences in reed shoot density (eCO2 vs. control plots: 195 ± 35 vs. 172 ± 21 in 2012, 213 ± 11 vs. 206 ± 11 in 2013; fertilization main effect ANOVA: p > 0.05; Figure 2B), mean plant height (eCO2 vs. control plots: 1.14 ± 0.70 m vs. 1.30 ± 0.82 m in 2012, 1.51 ± 0.91 m vs. 1.46 ± 0.88 m in 2013; main effects ANOVA: p > 0.05; Figure 2C), LAI (Figure 2D), and Abiomass (Figure 2E). Although Abiomass in the eCO2 plots was greater than in the control plots for most of the 2013 fertilization cycle, these differences were not significant (factorial ANOVA plot type × year; Bonferroni test: p > 0.05). The differences in maximum annual LAI and Abiomass in the eCO2 and control plots were not statistically significant (factorial ANOVA plot type × year: p > 0.05; Figure 3A, B). In addition, the increases in maximum annual LAI and Abiomass in 2013 were not significantly different (Figure 3A, B; factorial ANOVA plot type × year: p > 0.05). The stimulatory effect of eCO2 on Abiomass was significant (pooling all data: R 2 = 0.64, p = 0.003; Figure 3E), and this effect was particularly strong during 2013 (R 2 = 0.75, p = 0.02). In 2012, this effect was not statistically significant (p = 0.73; Figure 3E). The observed lag effect indicates that eCO2 stimulates growth when plant Abiomass is high, but growth is suppressed or null when Abiomass is low.

Figure 2
figure 2

Changes over time in water level (A), number of shoots (B), mean plant height (C), leaf area index (D), and above-ground biomass (E) of Phragmites australis exposed to eCO2 and control (ambient) conditions during 2012 and 2013. Each point indicates average and standard deviation (n = 6).

Full size image
Figure 3
figure 3

Maximum annual value (± SD) of leaf area index (A), above-ground biomass (Abiomass; B), and the effect of eCO2 on above-ground biomass in all six FACE plots for 2012 and 2013 (differences on Abiomass between reeds growing at eCO2 and those at ambient CO2). Each point is the maximum Abiomass in one treatment in each year. The dashed line is the correlation using all data (2012–2013: \( {\text{Abiomass}}_{{ + {\text{CO}}_{2} - {\text{AmbCO}}_{2} }} = 0.92 \times {\text{Abiomass}}_{{{\text{AmbCO}}_{2} }} {-}945 \), R 2 = 0.64, p = 0.002), and the solid line is the correlation using 2013 data (\( {\text{Abiomass}}_{{ + {\text{CO}}_{2} - {\text{AmbCO}}_{2} }} = 1.07 \times {\text{Abiomass}}_{{{\text{AmbCO}}_{2} }} {-}1320 \), R 2 = 0.74, p = 0.03). All factorial ANOVA plot type × year results (in A, B) were not statistically significant (p > 0.05).

Full size image

Stomatal Conductance, Leaf-Level Transpiration, Photosynthesis, and ITE

The temporal patterns of stomatal conductance, transpiration, and photosynthesis rate were very similar in the eCO2 and control plots during both fertilization cycles (Figure 4A–C; factorial ANOVA plot type × month: p > 0.05). However, the patterns of changes in stomatal conductance and photosynthesis rate were different in 2012 and 2013 (Figure 4A, C), but the transpiration rate had a similar pattern in 2012 and 2013 (Figure 4B). The temporal patterns of ITE were similar during each year, with lower values during mid-summer (Figure 4D). Although ANOVA indicated the fertilization effect was significant for both years (df = 2, F = 38.93, p < 0.0001), the interaction between plot type (eCO2 vs. control) × month (factorial ANOVA) was significant only for ITE values among plots for October 2013 (Bonferroni post hoc test: p = 0.009). The annual cumulative l transpiration and photosynthesis rates were significantly lower in all plots during 2013 (Figure 5A, B; main effect ANOVA: p = 0.004 and 0.001, respectively); however, differences in transpiration and photosynthesis between the eCO2 and control plots were not significant (factorial ANOVA plot type × year: p > 0.05; Figure 5A, B). Differences between transpiration and photosynthesis rates recorded during the early growth period (June) and during the senescent period (October) were significant during 2013 (Bonferroni post hoc test: p = 0.0001 and 0.002, respectively; Figure 5C, D).

Figure 4
figure 4

Changes over time in stomatal conductance (A), transpiration rate (B), photosynthesis rate (C), and instantaneous transpiration efficiency (ITE) (D) of Phragmites australis exposed to eCO2 and control (ambient conditions) during 2012 and 2013.

Full size image
Figure 5
figure 5

Annual cumulative transpiration (A) and photosynthesis (B) and average transpiration rate (C) and photosynthesis rate (D), recorded in June and October in Phragmites australis in plots growing Phragmites australis exposed to eCO2 and control (ambient) levels of CO2 during 2012 and 2013. Different letters above the bars indicate significant differences (Bonferroni test: p < 0.05). **p < 0.0001 in the Bonferroni post hoc test of the factorial ANOVA plot type × month.

Full size image

Relationships Between Transpiration, ITE, and Above-Ground Biomass

We examined monthly averaged data from each experimental plot (June, July, August, September, and October) and found no significant relationship between transpiration rate and ITE with monthly Abiomass (p > 0.05). However, there were significant relationships between transpiration and Abiomass when eCO2 and control data were grouped across plots using monthly averaged data (Figure 6A). The Abiomass response in the eCO2 and control plots had the same pattern during 2012 (p = 0.02), but in 2013, there was a greater increase in Abiomass in the eCO2 plots than the control plots, with similar transpiration rates (p = 0.04 and p < 0.0001, respectively; Figure 6A). We found no significant relationship between ITE and Abiomass using the grouped eCO2 and control monthly averages (p > 0.05, Figure 6B). Because eCO2 has a cumulative effect on plant growth, ITE values in September were inversely correlated with Abiomass (Figure 7). Analysis of all data (eCO2 + control) indicated a significant correlation only during 2012 (ITE = −10−6 × Abiomass + 0.0044, p = 0.02; Figure 7A). Analysis of ITE values from the eCO2 plots indicated different responses in 2012 and 2013 (Figure 7B). An ITE decrease, along with an Abiomass increase, declined considerably in 2013 for plants exposure to eCO2 (65 vs. 16%; Figure 7B). Plants growing under ambient CO2 had a declining in ITE response, which was related to functional changes (Figure 7C).

Figure 6
figure 6

Pearson correlations between monthly averages of biomass and transpiration rate (A) and instantaneous transpiration efficiency (ITE; B) in plots of Phragmites australis exposed to eCO2 and control (ambient) conditions during 2012 and 2013. Error bars represent standard deviations. For biomass, the variability associated with the prediction models was included within standard errors (see “Materials and Methods”).

Full size image
Figure 7
figure 7

Pearson correlations between above-ground biomass (Abiomass) and instantaneous transpiration efficiency (ITE) in plots exposed to eCO2 and control (ambient) conditions, using data from September of 2012 and 2013. A Regression for ITE values of eCO2 and control plots in 2012 (p = 0.02) and 2013 (p > 0.05); B regression for eCO2 plots in 2012: ITE = −10−6 × Abiomass + 0.0046, p = 0.04; regression for eCO2 plots in 2013: ITE = −2 × 10−6 × Abiomass + 0.0037, p = 0.04; C regression line for 2012 and 2013 together in controls plots: ITE = 0.025 × Abiomass−0.31, p = 0.01.

Full size image

Changes in Chl (a + b) and in Leaf and Soil Carbon, Nitrogen, and Phosphorus

Plants grown under eCO2 had similar chl (a + b) levels as those grown under ambient conditions during both years (factorial ANOVA plot type × month: p > 0.05; Figure 8A). In addition, the Chl (a + b) content increased in the eCO2 plots throughout the growing period (June–October) of both study years, and this increase was significant during 2012 (factorial ANOVA plot type × month between June and October: Bonferroni test: p = 0.03; Figure 8A). Despite this difference, the chl (a + b) levels in the control and eCO2 plots were similar by the end of the growing season (Figure 8A).

Figure 8
figure 8

Average levels of chlorophyll (a + b) (A), carbon (B), nitrogen (C), and phosphorus (D) in Phragmites australis leaves and organic carbon (E), organic nitrogen (F), and total P (G) in soil at the beginning (June) and the end (October) of the growing seasons in plants exposed to eCO2 and control (ambient) conditions during 2012 and 2013. **p < 0.05, in a Bonferroni post hoc test of the factorial ANOVA plot type × month.

Full size image

The leaf carbon content of plants exposed to eCO2 decreased significantly from June to October during both years (factorial ANOVA plot type × month between June and October: Bonferroni post hoc test in 2012 and 2013: p < 0.0001; Figure 8B) and also decreased significantly in control plants during 2012 (Bonferroni post hoc test: p < 0.0001; Figure 8B). The leaf declines in nitrogen were not significant in the eCO2 plots (Figure 8C). The changes in leaf phosphorus had different patterns in 2012 and 2013, but none of these changes were significant in the eCO2 plots (Figure 8D). Analysis of the carbon/nitrogen and nitrogen/phosphorus ratios of reed leaf tissues indicated no significant differences between eCO2 and control plots during both years (Table 2). During 2012, the carbon/nitrogen ratios decreased slightly, whereas the decline of the nitrogen/phosphorus ratio was more pronounced (≈ 36%) with respect to 2013. In contrast, the carbon/nitrogen and nitrogen/phosphorus ratios both increased during 2013 (Table 2).

Table 2 Mean Molar C/N/P Ratios (5 Samples Per Plot, 6 Plots Per Treatment (Mean ± Standard Deviation) of Phragmites australis Leaf Tissue Growing Under eCO2 and Control (Ambient) Conditions at the Beginning and the End of the Growing Seasons of 2012 and 2013.
Full size table

The soil organic carbon content increased significantly in the eCO2 plots (factorial ANOVA plot type × month between June and October: Bonferroni post hoc test in 2012: p = 0.002 and in 2013: p = 0.02; Figure 8E). However, these increases were not significant in the control plots (Bonferroni post hoc test: p > 0.50 for both years; Figure 8E). During the second year, the increase in soil organic carbon was significantly higher in the eCO2 plots than the control plots (factorial ANOVA plot type × month in October: Bonferroni post hoc test: p = 0.017). The temporal increases in soil organic nitrogen were also significant in all plot types (factorial ANOVA plot type × month between June and October; Bonferroni post hoc test in eCO2 plots: p < 0.005 and in control plots: p < 0.007), although the eCO2 and control plots were not significantly different (factorial ANOVA plot type × month: Bonferroni post hoc test: p > 0.5; Figure 8F). Soil phosphorus decreased consistently during both experimental cycles in the eCO2 and control plots, but was only statistically significant in the eCO2 plots during 2012 (factorial ANOVA plot type × month between June and October: Bonferroni post hoc test: p = 0.008; Figure 8G).

Discussion

Our results show that the effects of eCO2 [CO2 ≈ 582 µmol mol−1] on a wetland environment were more complex than previously observed. In particular, eCO2 generated a complex network of interactions among different environmental parameters that affected physiological processes (for example, photosynthesis, respiration, carbon allocation, plant nutrition, stomata function, and transpiration) related to plant age and adaptation to the new conditions. Our findings showed that eCO2 began to have some physiological effects on Phragmites during the second year, but did not significantly alter growth or biomass. The physiological effects of eCO2 are cumulative, because they appear at the end of the growing season. In contrast, growth chamber experiments using eCO2 reported an initial and immediate stimulation of macrophyte growth in numerous species (for example, S. olneyi, S. patens, Typha latifolia, T. angustifolia, Spartina maritima, and P. australis; Drake and others 1997; Drake and Rasse 2003; Rasse and others 2005; Sullivan and others 2010; Mateos-Naranjo and others 2010; Mozdzer and Megonigal 2012). Thus, it seems inappropriate to extrapolate the results of growth chamber experiments to predict the overall behavior of wetlands and other complex ecosystems to eCO2. This conclusion is supported by a meta-analysis of FACE experiments (Ainsworth and Long 2005), which showed that eCO2 stimulation of plant growth in growth chambers may be greater than in nature because factors that affect plant growth (for example, high temperature, elevated ground-level ozone, and changes in soil moisture) are not allowed to vary. In fact, eCO2 can reduce plant growth when combined with other environmental stressors (Shaw and others 2002). It follows that ecosystem responses to global climate change strongly depend on the interactions of many factors and that responses can be synergistic, antagonistic, or neutral (Martín and others 2014).

Environmental and Experimental Factors Influencing Responses of Wetland Plants to eCO2

The response of reed growth to eCO2 in our FACE experiment highlights the importance of complex interactions of factors under in situ field conditions. We aimed to study the effect of eCO2 under the natural ecological conditions of the wetland (including natural variability of abiotic and biotic patterns and processes; Sánchez-Carrillo and Angeler 2010), but our experiment still had some limitations. In particular, it is difficult to identify the causal mechanisms responsible for the observed plant responses because of the broad scale of factors that might have influenced ecosystem responses. In fact, we acknowledge that certain broad-scale factors that occurred after and during our experimental period (that is, severe drought followed by an above-average water, with prolonged flooding) may have affected some of our observations under eCO2. In fact, the lack of treatment effects on all reed plants could be because there was greater waterlogging of the root zone during most of the growing season, with limited phosphorus availability (see below). It is also necessary to consider phenomena associated with these hydrological changes that also occurred during the experimental period (a long aerobic period before the experimentation, high external and internal nutrient loading during flooding, availability of oxidized soil organic matter, and shifts in soil microbial composition), which also could have influenced plant responses to eCO2. Therefore, the recovery of the wetland from disturbance (including the preparation of the area for experimentation) could have affected our results.

Also, the specifics of our experimental design might have affected some of the results. In particular, rhizome planting and early growth of the plants may also have affected their responses to eCO2. Plant age may have affected the response to eCO2, as shown by changes in the Abiomass results from 2012 to 2013. P. australis exhibits a distinctive growth dynamic during its early stages, in that after the first year, there is a decline in the use of non-structural carbohydrates stored in the rhizomes from the previous growing season (that is, the shoots become more self-supporting; Hara and others 1993). Tracking plant growth over a longer time period would be important to capture these lag effects, measuring, for example, the amount of carbohydrates stored from the previous growing season. In agreement, we identified an increase in the variability of LAI and Abiomass between plots during the second year. Moreover, plant age also affects tolerance to deep water (Weisner and Ekstam 1993). All these factors are part of the complex wetland response to eCO2 that would not be observed in growth chamber studies. Our observations therefore highlight the importance of considering the interactions of multiple factors, rather than simply increased carbon uptake from photosynthesis, in determining the response of wetlands to eCO2.

Variable Hydrology and Wetland Response to eCO2

The variable hydrology of wetlands can induce complex ecosystem responses to eCO2. For example, flooding increases external nutrient inputs in seasonal wetlands and modifies nutrient transformations due to oxygen depletion of soils, and this increases production of end-products that are potentially toxic or interfere with nitrogen assimilation (Koch and others 1990; Megonigal and others 2005). Also, different adaptations that allow flooding-tolerance of emergent macrophytes (which affect the oxygen uptake pathway by roots and other physiological mechanisms; Kozlowski and Pallardy 2002) and competition can affect the distribution of macrophyte species in wetlands (Spence 1982; Van der Valk and others 1994; Sánchez-Carrillo and others 2004). Water depth and its seasonal variability alter growth in reed, reflecting its phenotypic plasticity (Vretare and others 2001), but can also affect transpiration, photosynthesis, and stomatal conductance (Brix and others 2001; Saltmarsh and others 2006). Thus, all of these factors can influence plant responses to eCO2. Transpiration and photosynthesis rates increase when reed plants are exposed to elevated water levels; intrinsic water use efficiency (ITE) and CO2 utilization also increase when plants are grown under water stress (Patger and others 2005) and in field conditions during a prolonged drought (Ortíz-Llorente 2013). Our findings show that, at least during the second year and in saturated soils, Phragmites had improved ITE under eCO2 relative to ambient CO2 at the end of the growing season. However, these improvements could not be related to a decline in stomatal conductance. In accordance with the phenotypic plasticity of reeds and responses observed in other grasses (Huxman and others 1998; Lecain and others 2003), we expect that reed responses to eCO2 might have been different under conditions of water limitation. In three large-scale FACE experiments with C4 species, Leakey and others (2009) showed that plants produced the same Abiomass and yield at ambient and eCO2 during a growing season free of drought, but there were significant differences under conditions of water stress. Therefore, it may be expected that differences of reeds grown under eCO2 and ambient CO2 would be more pronounced under conditions of water stress. The differences of our results with those from growth chamber experiments (Mozdzer and Megonigal 2012) can be partially explained by the low water level (≈ 3 cm) present during their entire experiment, which could be considered a water stress scenario.

Although eCO2 had no significant effect on stomatal conductance in Phragmites, our observed changes in photosynthesis are in accordance with the optimal stomatal behavior defined by Cowan and Farquhar (1977), which is related to ITE. Therefore, ITE should increase in proportion to atmospheric CO2 when plants are exposed to eCO2 (Barton and others 2012). In our experiments, the 46% increase in atmospheric CO2 (from 399 to 583 µmol mol−1) and the 38–44% increase in average ITE were in accordance with the reported performance of this FACE (Sánchez-Carrillo and others 2015). Based on the theory of optimal stomatal behavior, our results predict a distinctive response of reed plants exposed to eCO2 that depends on water availability (or hydroperiod): during wet periods reed enhances stomatal conductance and carbon uptake increases regardless of atmospheric CO2 concentration; contrarily, dry cycles induce stomatal closure in reed, but this reduction is lower in plants exposed to eCO2, increasing significantly carbon uptake compared with those growing at current CO2 concentrations. If these dynamics predict wetland behavior at the ecosystem-scale, then the responses of emergent macrophytes to long-term eCO2 could be even more complex, given the relatively large seasonal and annual fluctuations in water level in most freshwater wetlands, particularly in semiarid zones.

Carbon Losses from Roots as a Response of Wetland Plants to eCO2

Probably our most interesting result is that there was increased soil organic carbon in plots exposed to eCO2. This suggests significant carbon losses from roots and may partly explain the lack of increased growth and carbon accumulation following eCO2. The ability of roots to secrete a vast array of compounds into the rhizosphere is one of their most remarkable metabolic features. Plants may release up to 20% of photosynthesis products into the soil, and this provides the basis for establishment of plant-microorganism interactions that benefit plant growth by increasing the availability of minerals, production of phytohormones, degradation of phytotoxic compounds, and suppression of soil-borne pathogens (Bais and others 2006). Some other studies reported that root exudation plays a major role in maintaining root–soil contact in the rhizosphere because it modifies the biochemical and physical properties of this environment due to fluctuations in hydration, and therefore contributes to root growth and plant survival (Walker and others 2003). Vascular species under eCO2 have increased carbon release due to enhanced root exudation of carbon and root turnover (Fenner and others 2007). When aquatic plants, such as Orontium aquaticum, E. fluviatile, and Phragmites communis, are exposed to eCO2, they also exhibit increased photosynthetic rates, but not Abiomass because of the increased root exudation of organic carbon (Megonigal and Schlesinger 1997; Ojala and others 2002; Kim and Kang 2008). Some FACE experiments in rice fields demonstrated that the increase in soil-labile carbon regulated microbial activity, which in turn increased mineral transformations (Guo and others 2012). It is likely that this was also a main effect of root exudation of carbon in our FACE experiment (unpublished results). It seems that water stress is responsible for the increased root exudation of carbon under eCO2, because similar responses were reported in several other aquatic plants, such as Vallisneria americana, Juncus bulbosus, P. australis, and T. latifolia, when they are in stressful conditions, including low level of light, increased acidity, and exposure to xenobiotics (Chabbi and others 2001; Kurtz and others 2003; Larue and others 2010). It remains to be determined whether the sudden change to eCO2 or flooding had the more significant effect of the reed plants. It is also unknown whether this plant response will be maintained as the plants age, another important issue related to the long-term dynamics of wetlands. These effects may be important in the context of wetland function, because they affect wetland soil biogeochemistry and the potential of wetlands to act as carbon sinks in a future with increased atmospheric CO2.

Conclusions

Experiments under controlled laboratory conditions in growth chambers have provided an important basic understanding of the responses of aquatic plants to rising CO2 and insight into the potential mechanisms of these responses. The FACE approach that we used here allows examination of aquatic plants under field conditions and can potentially provide environmentally relevant impact assessments. Our results showed that response of P. australis and the wetland environment to eCO2 is more complex than previously observed, because specific hydrological conditions modulate these responses. Our results show that under eCO2, reed plants modify carbon allocation by increased root exudation of carbon, and this limits the increase in above-ground biomass. In addition, our findings indicate that water availability affects the improved ITE of reed plants following eCO2. Therefore, these combined effects of water stress and eCO2 on wetlands have implications for the function of wetlands as a carbon sink and for wetland hydrology. This means that global climate change could affect regional and global carbon and water cycles in wetlands. This study also highlights the combined effect of eCO2 with other stressors, which can modulate outcomes or even produce unexpected outcomes. Wetlands are complex transitional ecosystems, intermediate between terrestrial and aquatic ecosystems, and multiple mechanisms and processes interact at different temporal and spatial scales. Therefore, considering that several factors affect plant responses under eCO2, more long-term FACE experiments in different inland wetlands could improve our knowledge of how eCO2 affects wetland ecological and biogeochemical processes, and how these affect the stability of ecosystems and their capacity to function as carbon sinks under conditions of climate change.