Elevated CO2 enhances biological contributions to elevation change in coastal wetlands by offsetting stressors associated with sea-level rise


*Correspondence author. E-mail: julia.cherry@ua.edu


  • 1Sea-level rise, one indirect consequence of increasing atmospheric CO2, poses a major challenge to long-term stability of coastal wetlands. An important question is whether direct effects of elevated CO2 on the capacity of marsh plants to accrete organic material and to maintain surface elevations outweigh indirect negative effects of stressors associated with sea-level rise (salinity and flooding).
  • 2In this study, we used a mesocosm approach to examine potential direct and indirect effects of atmospheric CO2 concentration, salinity and flooding on elevation change in a brackish marsh community dominated by a C3 species, Schoenoplectus americanus, and a C4 grass, Spartina patens. This experimental design permitted identification of mechanisms and their role in controlling elevation change, and the development of models that can be tested in the field.
  • 3To test hypotheses related to CO2 and sea-level rise, we used conventional anova procedures in conjunction with structural equation modelling (SEM). SEM explained 78% of the variability in elevation change and showed the direct, positive effect of S. americanus production on elevation. The SEM indicated that C3 plant response was influenced by interactive effects between CO2 and salinity on plant growth, not a direct CO2 fertilization effect. Elevated CO2 ameliorated negative effects of salinity on S. americanus and enhanced biomass contribution to elevation.
  • 4The positive relationship between S. americanus production and elevation change can be explained by shoot-base expansion under elevated CO2 conditions, which led to vertical soil displacement. While the response of this species may differ under other environmental conditions, shoot-base expansion and the general contribution of C3 plant production to elevation change may be an important mechanism contributing to soil expansion and elevation gain in other coastal wetlands.
  • 5Synthesis. Our results revealed previously unrecognized interactions and mechanisms contributing to marsh elevation change, including amelioration of salt stress by elevated CO2 and the importance of plant production and shoot-base expansion for elevation gain. Identification of biological processes contributing to elevation change is an important first step in developing comprehensive models that permit more accurate predictions of whether coastal marshes will persist with continued sea-level rise or become submerged.


Increases in atmospheric CO2 may affect coastal plant communities directly by altering production via a fertilization effect, or indirectly by contributing to sea-level rise via global warming (IPCC 2007). It is unclear, however, if stressors associated with sea-level rise (flooding and salinity) will limit the enhancement effect of atmospheric CO2, and thus long-term stability of coastal marshes.

Tidal marshes maintain surface elevations relative to sea level through physical and biological processes (Fig. 1). Physical processes of sedimentation, erosion and subsidence control marsh elevations by influencing mineral sediment accretion and compaction, but biotic processes also affect soil volume, thereby contributing to elevation change (Nyman et al. 1993; Cahoon et al. 2006; McKee et al. 2007). Biogenic contributions to soil formation, which are controlled by production–decomposition dynamics, may be one mechanism by which coastal marshes adjust to prevailing water levels. In systems once reliant on sediment deposition, biological processes may be more important for long-term stability of coastal marshes than previously recognized. Thus, any changes in factors that affect plant production or decomposition, including elevated atmospheric CO2 concentrations, may affect the capacity of the marsh to avoid submergence.

Figure 1.

Summary of processes controlling elevation change in brackish marshes. Average values of above- and below-ground production are based on field measurements (J.A. Cherry & K.L. McKee, unpublished data).

Global forcing factors, such as eustatic sea-level rise and elevated atmospheric CO2, may directly or indirectly influence production and decomposition of marsh plants, and ultimately, surface elevations. Under elevated CO2 conditions, plant shoot densities and productivity can be enhanced in both the short-term (Drake et al. 1997; Reich et al. 2001; McKee & Rooth 2008) and long-term (Niklaus et al. 2001; Rasse et al. 2005; Erickson et al. 2007). Elevated CO2 also can contribute to increased below-ground inputs by enhancing root growth (Cheng & Johnson 1998; McKee & Rooth 2008).

Response to CO2 is often greater in C3 than in C4 species (e.g., Curtis et al. 1989, 1990) because CO2 enrichment reduces diffusion limitation and photorespiration in C3 species whereas photosynthesis is nearly saturated at current CO2 levels in C4 species (Woodward et al. 1991). Different responses to elevated CO2 by C3 and C4 species may lead to shifts in species composition (Polley et al. 2003) and losses in biodiversity (Niklaus et al. 2001; Reich et al. 2001). Elevated CO2 also may indirectly affect organic matter accumulation in marshes through changes in litter quality and decomposition rates (Curtis et al. 1990; Ball & Drake 1997; but see Norby et al. 2001). These effects may interact with increased flooding and salinity associated with sea-level rise (Erickson et al. 2007) or with local factors such as nutrient loading (McKee & Rooth 2008) to further modify above- and below-ground organic matter accumulation.

Warmer global-mean surface temperatures correlated with elevated CO2 concentrations contribute to sea-level rise as ocean waters undergo thermal expansion and polar ice caps melt (IPCC 2007; Rahmstorf 2007). As sea levels continue to rise, plants in coastal wetlands may become stressed by increased salt water intrusion and inundation (Scavia et al. 2002), which may contribute further to changes in production and decomposition rates, and thus biogenic accretion. Elevated CO2 levels may alleviate stress associated with increased salinity or flooding (Rozema et al. 1991; Lenssen et al. 1993; Rasse et al. 2005), and thereby influence relative rates of plant production. Rates of decomposition also may vary as flooding and salinity contribute to reducing conditions in the soil and introduce alternate oxidants (e.g., sulphate) that drive redox reactions (Rozema et al. 1991).

Many studies of climate change effects on plant production and decomposition have manipulated only one or two factors at a time (Curtis et al. 1989; Ball & Drake 1997; Reich et al. 2001; Derner et al. 2003). In addition, studies of elevation dynamics in coastal marshes typically have not considered biogenic (especially below-ground) contributions to elevation change, instead focusing more on surface deposition of mineral and organic matter, shrink-swell due to subsurface hydrology, and other physically-mediated controls on elevation (Morris et al. 2002; Krauss et al. 2003; Whelan et al. 2005; but see Rybczyk & Cahoon 2002 and McKee et al. 2007). To accurately predict the persistence of coastal wetlands, we need more information about multiple-factor interactions and direct and indirect controls on marsh elevations. More specifically, manipulative experiments examining interactions of CO2 and other factors are required to fully assess biological processes influencing accretion and elevation change.

Here, we test the hypothesis that elevated CO2, salinity, and flooding interactively drive changes in elevation of a brackish marsh community by differentially influencing C3 and C4 plant production. We conducted a glasshouse mesocosm experiment in which conditions could be closely controlled, physical processes (e.g., sedimentation) did not contribute to elevation change, and biological mechanisms influencing elevation change could be identified. Through structural equation modelling (SEM), we statistically evaluated multivariate hypotheses and identified direct and indirect pathways driving changes in surface elevations. This experimental design permitted evaluation of how CO2 enhancement might be modified by stress responses to increased salinity and flooding. The goal of this research was to understand linkages and feedback effects controlling soil elevations and how external forcing functions interact with internal processes to allow marshes to keep pace with sea-level rise. The results will lead to improved models to predict vulnerability of marshes under global change scenarios and to inform ongoing field studies of coastal marsh stability.


experimental design

Intact sods of soil and vegetation were collected from a brackish marsh in Big Branch Marsh National Wildlife Refuge, Louisiana, USA (N30°15.90′, W89°57.07′) in March 2005. At this site, a complementary study is examining biophysical controls on marsh elevation, which provides field observations on elevation change and plant production for comparison with controlled mesocosm experiments. Sods (c. 25 cm deep × 28.5 cm in diameter) consisted of approximately 50 : 50 mixture of Spartina patens (Ait.) Muhl. (marshhay cordgrass), a C4 species, and Schoenoplectus americanus (Pers.) Volk. Ex Schinz & R. Keller (American bulrush), a C3 species. Upon removal from the marsh, sods were placed in high-density polyethylene containers (20 L; 32 cm deep × 28.5 cm diameter) (hereafter mesocosms) and transported to the Wetland Elevated CO2 Experimental Facility at the National Wetlands Research Centre in Lafayette, LA, USA (facility described in McKee & Rooth 2008). Sods were trimmed from the bottom to a depth of 20 cm, a depth that captured the majority of the living root zone. Sods were repositioned in mesocosms atop a 2 cm layer of gravel (to facilitate drainage of water) so that the soil surface was 10 cm below the mesocosm rim. All above-ground biomass was removed to control for initial differences in vegetation cover.

The study used a split-plot design according to the prescription of Potvin (2000). In this design, four glasshouses served as whole plots with the main effect, CO2, being applied at the whole plot level. Two glasshouses were randomly assigned to both elevated and ambient CO2 conditions a priori, and suplot factors of flooding level and salinity were varied within glasshouses. Overall, each mesocosm was randomly assigned to one of 30 treatment combinations in which CO2 (ambient = 380 ppm and elevated = 720 ppm), salinity (0, 5, 10, 15, and 20‰ sea salts), and flooding (drained, intermittently flooded, and flooded) were manipulated. Each treatment combination was replicated twice (n = 60) and blocked by glasshouse (n = 15 per glasshouse). Ambient CO2, low salinity, intermittently flooded conditions were representative of conditions observed in situ.

CO2 concentrations, salinity and water levels were established and maintained for the duration of the experiment. CO2 concentration in each glasshouse was monitored by dual-beam, steady-state infrared gas analyzer (Gascard II, Edinburgh Instruments, Ltd., Livingston, UK) and maintained through automated feedback regulation of flow meters (Cole-Parmer Instrument Company, Vernon Hills, IL) supplying CO2 gas to glasshouse atmospheres (see McKee & Rooth 2008 for details). Air within the plant canopy was continually sampled in all glasshouses at 5 min intervals throughout the experiment, and feedback controllers maintained CO2 concentrations within 5% of the setpoint. Salinity treatments were achieved using Instant Ocean (Aquarium Systems USA). Water levels were created by drainage controls on each mesocosm and manually adjusted to provide unflooded (completely drained, but flushed once per day with water to prevent desiccation), intermittently flooded (alternating fully flooded and fully drained conditions on a monthly basis), and constantly flooded conditions (flooded 10 cm above original soil surface). Water was drained into individual reservoirs between flushings or during water exchanges. For intermittently flooded replicates, monthly duration of flooding was determined from the average of the previous 10 years of tide gauge data (USGS Lake Pontchartrain gauge) and the percentage of each month that water levels were greater than mean water level. On average, mesocosms were flooded 5.9 ± 1.7 days per month. This method allowed capture of seasonal and monthly variation in the hydroperiod, but not daily variations (note, however, that tides in this region are diurnal, microtidal, and influenced more by meteorological than astronomical factors).

Treatments levels were established by systematically increasing CO2 (45 ppm per week for 8 week), salinity (5‰ per week for 4 week), and flooding (specific to shoot height within mesocosms) to permit acclimation by plants to treatment levels (Klironomos et al. 2005). All treatment levels were achieved by 23 May 2005. Thereafter, mesocosms underwent water exchanges monthly and were adjusted as needed to maintain salinity. Water levels were checked daily and adjusted as needed to maintain flooded conditions. Mesocosms were amended twice per month with a nutrient solution providing mmol L−1 of N (0.005) as NH4Cl, P and K (0.00125) as KH2PO4, S (0.0025) as MgSO4, Ca (0.00625) as CaCl, Fe (0.00125) as FeEDTA solution, and micronutrients (0.00125). Plant shoots were periodically sprayed with DI water to remove accumulated salts. The experiment was terminated in May 2006 once significant treatment effects on elevation had been detected.

hydro-edaphic, plant, and elevation analyses

To characterize hydro-edaphic conditions within mesocosms and to maintain salinity treatments, soil redox potentials and porewater salinity, pH, sulphide, and bioavailable N and P (inline image and inline image) were measured quarterly from May 2005 to May 2006. Soil redox potentials (Eh) at 5 and 15 cm depths were determined using platinum electrodes equilibrated for 30 min (McKee et al. 1988). Porewater was extracted at a depth of 15 cm, and salinity, pH, and sulphide were determined as described in McKee et al. (1988). Porewater extracted for nutrients was analyzed for inline image and inline image (Robertson et al. 1999) using a LACHAT system (Quikchem IV, Lachat Instruments, Milwaukee, WI).

Shoot densities of each species were counted quarterly from May 2005 to May 2006. To determine above-ground production, senescent shoot material was collected monthly, separated to species, oven-dried at 60 °C to a constant mass, and weighed (±0.01 g). Thus, senescent material did not contribute litter to the soil surface of mesocosms, thereby isolating root zone from litter contributions to elevation change. At the end of the experiment, all above-ground material was harvested, separated by species into live and dead components, oven-dried at 60 °C to a constant mass, and weighed (±0.01 g). Above-ground net primary production for each species was calculated by summing monthly senescent masses and final harvested mass. Below-ground biomass accumulation was determined using root ingrowth bags (Symbula & Day 1988; Neill 1992), which provided a relative measure of root and rhizome contribution to soil volume. This approach measured the net result of root production, turnover, and decomposition between measurement intervals and provided a relative estimate of root mass accumulation across treatments and, consequently, related directly to the dependent variable of interest: elevation change. Ingrowth bags, constructed of a flexible mesh (5 mm) filled with milled sphagnum peat, were inserted into a hole (5 cm diameter × c. 20 cm deep) cored from a randomly selected quadrant in each mesocosm. Unamended peat was similar in bulk density and organic content to the native substrate, but lower in nitrogen content. Ingrowth bags were collected three times during the experiment in August 2005, November 2005 and May 2006. Sampling at this frequency maximized quantification of root contributions, but minimized quantification of rhizome contributions to below-ground biomass accumulation (rhizomes found in < 50% of samples). All in-grown below-ground material was washed over a 1-mm mesh sieve, oven-dried at 60 °C to a constant mass, and weighed (±0.0001 g). In-grown root material was not separated by species due to the difficulty of identification; consequently, results represent community accumulation of below-ground root matter. Annual below-ground accumulation of root mass was calculated separately by summing all biomass produced at intervals throughout the year. Because rhizome contributions were minimized and did not contribute to elevation change (R2 = 0.01, F1,58 = 0.35, P = 0.56), only the accumulation of roots was considered during SEM analyses.

Soil surface elevation was measured quarterly from May 2005 to May 2006 using a miniature surface elevation table (mini-SET) designed after the rod-SET used to measure elevation change in situ (Cahoon et al. 2002). The mini-SET consisted of a removable measuring arm anchored to the mesocosm rim in one of two positions. In each position, 13 fibreglass pins (n = 26 per mesocosm) were lowered to the soil surface through holes in the arm, and pin extension distances (mm) above the arm were measured. Changes in pin extension distances over time corresponded to changes in soil surface elevation. Elevation change corresponded to the total change in sod elevations (mm) between May 2005 and May 2006. Determination of the soil surface was performed by the same person throughout the experiment based on a pre-established definition that excluded any unincorporated material or pins that intercepted plant shoots. Thus, measured responses of elevation change were based on the same criteria regardless of treatments.

After the final harvest, a core (7 cm diameter × c. 20 cm deep) was removed from the centre of each mesocosm to assess the biomass and volume of below-ground shoot bases, which were defined as portions of the plant shoots or stems located underground between the rhizome and soil surface. Shoot bases were assessed separately because they were not included in either above- or below-ground measures of annual production (described above). Because these cores contained some biomass produced before treatment initiation, they were not included in the SEM. Instead, these data were obtained to calculate potential contribution of shoot bases to soil displacement and to aid in interpreting biological inputs to elevation change identified by the SEM. All below-ground material in the core was washed over a 1 mm mesh sieve. Shoot bases were separated by species, attached rhizomes and roots were removed, shoot-base size (e.g., diameter) and culm density were recorded, and volume was determined by water displacement. Shoot-base volume per mesocosm was calculated for each species (based on culm density) and then correlated with total elevation change.

statistical analyses

Analysis of variance

Net responses in elevation change were analyzed using conventional anova procedures. Since our design followed the prescription of Potvin (2000), we also followed her prescription for analysis, using split-plot GLM and testing main effects using the whole-plot error term (using SAS version 8, SAS Institute, Inc, Cary, NC). The model tested for the effect of glasshouses, including potential differences in humidity, temperature, and shading. Data were screened prior to analyses; no indications were found to suggest the need for data transformations.

Structural equation modelling (SEM)

We used SEM (Grace & Pugesek 1997; Grace 2006) to evaluate hypotheses about processes underlying treatment influences on elevation change. SEM provides a means of evaluating path models that specify how exogenous factors (in this case, treatments) lead to system responses. We began by specifying a meta-model (Fig. 2), which represents the relationships of interest in the analysis. Here, our goal was to evaluate effects of CO2, salinity, and flooding on production by C3 and C4 plants, and how these processes ultimately lead to changes in surface elevation. The analysis provides both a quantification of the strengths of relationships and an evaluation of adequacy of model structures (e.g., are there paths missing, are some paths not required to explain the data?). In this case, we focused on several specific questions:

Figure 2.

Conceptual construct model presenting hypothesized direct and indirect effects of treatments (shaded boxes) on biotic variables and elevation change.

  • 1To what degree does elevated CO2 enhance production by the C3 species? Also, does CO2 interact with effects of salinity and flooding?
  • 2Do salinity and flooding stress have differential effects on production by the C3 and C4 species?
  • 3Do C3 and C4 species show evidence of competitive suppression or enhancement of each other?
  • 4How does root production relate to above-ground production by C3 and C4 species and how is this modified by hydro-edaphic stressors?
  • 5To what degree does change in surface elevation relate to total production responses by C3 and C4 species and how is this influenced by hydro-edaphic stressors?
  • 6Are there other processes operating whereby CO2 and hydro-edaphic stressors influence elevation change?

Using Mplus (Muthén & Muthén 2003), we formulated and estimated statistical models with the form in Fig. 2. As a first stage of analysis, distributional assumptions were examined for all variables to determine if transformations were necessary. Additionally, bivariate relationships were examined for indications of nonlinearities before inclusion in the SEM. Incorporated into the SEM analysis process was a consideration of multivariate nonlinearities, including interactive effects. Since initial analysis indicated that root production did not contribute to the explanation of elevation change, this variable was dropped from subsequent model analyses. Model adequacy was evaluated based on model χ2-values and associated P-values. Note that model adequacy is indicated by P-values > 0.05, because the goal of SEM is to evaluate the theoretically-specified models, not null models.

Additional analyses

Based on SEM results, additional analyses were conducted to further our understanding of pathways contributing to elevation change in this experiment. Because methods used to quantify above- and below-ground production could not adequately capture the contribution of shoot bases to soil expansion, final shoot-base volume was assessed separately at the end of the experiment (described above). Shoot-base volume was then correlated to elevation change. All correlation and regression analyses were performed using JMP 6.0.3 (SAS Institute, Inc, Cary, NC).


hydro-edaphic conditions

Soil redox potential (Eh), porewater sulphide concentrations, and porewater pH were measured to characterize hydro-edaphic conditions in mesocosms during the experiment. At final harvest, mean Eh indicated strongly reducing to oxidizing conditions at both 5 cm and 15 cm depths, which reflected flooding treatments (flooding effect: F2,28 = 66.12 and 61.81, P < 0.001 at 5 and 15 cm, respectively) (Table 1). Sulphide concentrations increased as flooding and salinity increased, although a significant three-way interaction indicated that the sulphide response varied with CO2 as well (F8,28 = 2.79, P = 0.02). Porewater pH tended to be higher in flooded treatments than in drained treatments, although this pattern was less apparent than for Eh and sulphide and also varied with salinity and CO2 (three-way interaction: F8,28 = 2.57, P = 0.03) (Table 1).

Table 1.  Hydro-edaphic conditions (mean ± 1SE) of treatment combinations at time of final harvest (n = 2 for each). Letters denote significant differences (P < 0.05) among pairwise comparisons for each edaphic variable. Dashes indicate inclusion of all letters between the two letters noted
CO2Treatments*Eh (mV)Sulphide (mM)pH
FloodingSalinity5 cm15 cm
  • *

    A, ambient; E, elevated; Dr, drained; Int, intermittently flooded; Fl, flooded.

  • nd, not detected/below detectable limits.

 Dr0427.0 ± 100.0a338.5 ± 22.5and4.51 ± 0.19fgh
5300.5 ± 142.5a417.5 ± 22.5a0.01 ± 0.00a4.16 ± 0.15gh
10447.0 ± 64.0a361.0 ± 64.0and4.50 ± 0.12fgh
15336.5 ± 28.5a415.0 ± 15.0a0.12 ± 0.11a4.69 ± 0.23e–h
20350.5 ± 83.5a318.5 ± 111.5and5.20 ± 0.07b–h
AInt0452.5 ± 90.5b188.0 ± 9.0a0.03 ± 0.02a5.31 ± 0.53a–h
5227.5 ± 51.5b225.0 ± 102.0a0.01 ± 0.00a5.36 ± 0.57a–h
10175.5 ± 49.5b220.5 ± 82.5a0.02 ± 0.01a5.92 ± 0.59a–f
15242.0 ± 27.0b280.5 ± 53.5a0.01 ± 0.00a6.42 ± 0.26a–c
20262.0 ± 78.0b233.5 ± 186.5a0.01 ± 0.00a6.71 ± 0.08ab
053.5 ± 49.5c–19.0 ± 5.0b0.14 ± 0.04a5.79 ± 0.31a–g
5–5.0 ± 39.0c–28.5 ± 6.5b0.26 ± 0.10a5.94 ± 0.06a–f
Fl108.0 ± 2.0c–75.5 ± 240.5b0.28 ± 0.11a5.89 ± 0.13a–f
155.5 ± 77.5c–24.5 ± 62.5b0.33 ± 0.30a5.78 ± 0.45a–g
20120 ± 60.0c–40.0 ± 129.0b2.41 ± 0.65b6.89 ± 0.09a
0334.0 ± 16.0a330.5 ± 107.5a0.01 ± 0.00a4.40 ± 0.45fgh
5479.0 ± 14.0a428.0 ± 30.0a0.01 ± 0.00a3.90 ± 0.20h
Dr10342.0 ± 107.0a340.5 ± 16.5and4.72 ± 0.02d–h
15403.5 ± 52.5a260.5 ± 40.5a0.01 ± 0.00a5.35 ± 0.49a–h
20299.0 ± 23.0a374.5 ± 48.5and5.48 ± 0.06a–h
0307.5 ± 178.5b358.5 ± 57.5a0.01 ± 0.00a5.68 ± 0.62a–g
5255.0 ± 52.0b325.0 ± 84.0a0.01 ± 0.00a6.22 ± 0.21a–e
EInt10328.5 ± 61.5b356.0 ± 6.0a0.03 ± 0.02a5.64 ± 0.49a–g
15357.0 ± 69.0b345.5 ± 106.5a0.25 ± 0.24a5.02 ± 0.13c–h
20262.0 ± 36.0b149.0 ± 164.0a0.02 ± 0.01a6.51 ± 0.12abc
0127.0 ± 5.0c2.0 ± 13.0b0.08 ± 0.02a5.99 ± 0.05a–f
5–29.0 ± 103.0c–88.5 ± 1.5b0.63 ± 0.10a6.23 ± 0.16a–e
Fl10–62.5 ± 62.5c–142.0 ± 141.0b0.65 ± 0.53a5.95 ± 0.39a–f
1550.5 ± 51.5c30.0 ± 90.0b0.49 ± 0.44a6.10 ± 0.38a–f
20150.0 ± 6.0c–128.5 ± 11.5b0.93 ± 0.07b6.38 ± 0.10a–d

anova results

Soil elevation change in mesocosms varied from a loss of –4.7 ± 2.3 to a gain of 44.0 ± 6.5 mm year−1, which was similar to observed field measurements of elevation change (–15.0 to 36.0 mm year−1) within the rooting zone at Big Branch Marsh NWR (Cherry & McKee, unpublished data). Soil elevation increased in both ambient and elevated CO2 conditions over the course of the experiment (0.79 ± 0.21 and 1.24 ± 0.24 mm year−1, respectively), although the rate of change was faster for elevated CO2 (F1,28 = 14.57, P = 0.0007) (Fig. 3). Total elevation gain was more than 50% greater at elevated CO2 (Table 2; Fig. 4). Main effects of water level and salinity on elevation change were also evident (Table 2), as elevation change decreased with increased flooding and exhibited a unimodal response to increasing salinity (Fig. 4). Significant two-way and three-way interactions (Table 2), however, indicated that elevation response to CO2 varied with flooding and salinity.

Figure 3.

Change in elevation in ambient (open circles) and elevated (shaded circles) CO2 treatments over time. Values are the mean ± 1 SE (n = 30).

Table 2. anova results for elevation change. Table abbreviations: source, source of variation; d.f., degrees of freedom; SS, sums of squares; MS, mean squares; F, F-value; P, probability under a null model; G, glasshouse effect, W, water level effect, S, salinity effect
Whole plot
 Whole plot error238.419.2  
 W27311.43655.7118.80< 0.0001
 S31435.4 358.811.70< 0.0001
 W × S81523.6190.56.19< 0.0001
 CO2 × W2225.2112.63.660.0387
 CO2 × S4291.973.02.370.0765
 CO2 × W × S8639.880.02.600.0292
 Subplot error28861.730.8  
Figure 4.

Main effects of ambient or elevated CO2 (top panel), flooding (middle panel), and salinity (bottom panel) on total elevation change. Elevation change refers to the total change in elevation over the course of the experiment (mm). Values are the mean ± 1 SE (n = 30, 20, and 12, respectively). Letters denote significant differences (P < 0.05) among pairwise comparisons.

Bivariate relationships indicated that elevation change was driven by biological processes (Fig. 5). Elevation change was positively correlated with above-ground production of S. americanus (r = 0.87, F1,58 = 174.87, P < 0.0001), but negatively correlated with S. patens production (r = –0.65, F1,58 = 41.72, P < 0.0001). Above-ground production of S. americanus and S. patens reflected species-specific stress tolerances of flooding and salinity (Fig. 6). Rhizome contributions represented 21.25 ± 3.23% of total below-ground production and did not contribute to elevation change (R2 = 0.01, F1,58 = 0.35, P = 0.56). Thus, rhizomes were excluded from measures of community below-ground production. Elevation gain was greater, however, when community below-ground production (root accumulation only) was greater (r = 0.39, F1,58 = 10.46, P < 0.002), although this relationship was weaker than those involving above-ground measures (Fig. 5). Instead, a strong positive correlation between final volume of shoot bases (below-ground storage structures) of S. americanus and elevation change (r = 0.84, P < 0.0001) suggested shoot bases were more important than roots in controlling elevation and that the C3 species was primarily driving soil expansion (Fig. 7). Elevation change was negatively correlated with shoot-base volume of S. patens (r = 0.44, P = 0.0006), which contributed less to soil volume than did S. americanus (Fig. 7).

Figure 5.

Bivariate relationships between specific biotic predictors and elevation change. C3 Prod refers to S. americanus above-ground production, C4 Prod refers to S. patens above-ground production, Community BG refers to below-ground root accumulation, and Elevation Change refers to the total change in elevation over the course of the experiment (mm).

Figure 6.

Relative stress tolerance of brackish marsh plant species as indicated by above-ground production of S. americanus (C3; open bars) and S. patens (C4; grey bars) to flooding (top panel) and salinity (bottom panel) (averaged over other treatments). Values are the mean ± 1 SE (n = 20 or 12, respectively).

Figure 7.

Relationship between elevation change and shoot-base volume of S. americanus (C3) and S. patens (C4). Values are total volume for each mesocosm based on average shoot-base volume multiplied by total shoot density per species.

sem results

The SEM model was found to have a close fit to data (χ2 = 11.56, d.f. = 8, P = 0.17) (Fig. 8). Model adequacy is indicated by P-values > 0.05, and in this case, metrics indicate a good fit between model and data. The model explained 82% of the variance in S. americanus production, 50% of the variance in S. patens production, and 78% of the variance in elevation change, further suggesting that the model was adequate as a means of representing relationships in the data.

Figure 8.

Structural equation model developed for experimental data to test for direct and indirect effects of CO2, salinity, and flooding (shaded boxes) on elevation change. Biotic response variables are (i) S. americanus production (C3 Prod), and (ii) S. patens production (C4 Prod). Solid arrows indicate significant pathways; squares along pathways (inline image) indicate interactions. Coefficients along pathways are standardized partial regression coefficients (r); negative mathematical signs indicate negative relationships; ns indicates non-significance; width of arrows represent strength of relationships as indicated by corresponding r values. R2 values are provided for dependent variables.

The hypothesized relationships of the construct model (Fig. 2) and the final results (Fig. 8) differed in several ways. We hypothesized that (i) CO2 would have a direct, positive effect on S. americanus (C3) production; (ii) salinity and flooding would directly influence both the C3 and C4 (S. patens) species; (iii) the two species would mutually reduce each other's growth; (iv) flooding and above-ground production of both species would affect root accumulation, and (v) flooding, and above- and below-ground production of both species would drive changes in elevation. SEM results indicated that several of these hypothesized relationships were not consistent with the data. First, there was not a significant direct pathway from CO2 to C3 production. Instead, the influence of CO2 in the model was to ameliorate salinity effects on S. americanus (C3) production (Fig. 7; path coefficient, r = –0.26). Second, coefficients linking salinity and flooding to plant responses were quite dissimilar. While S. americanus production was lower at higher salinities, S. patens production did not vary with salinity. Schoenoplectus americanus production was higher at higher levels of flooding, while S. patens production was strongly lower. Third, the SEM suggests an effect by S. patens on S. americanus, with no indication of a reciprocal effect. Fourth, community root production did not contribute to explaining variations in elevation change (P > 0.05) and thus, this variable was dropped from the final model for simplicity. Fifth, elevation change was found to be most strongly associated with production by the C3 species (S. americanus), even though this species’ production was also significantly enhanced by flooding. Finally, no evidence of direct pathways from COor salinity to elevation change was found, nor were any other interactive effects found, including the three-way interaction among CO2, salinity and water level observed in the anova results (probably because of the explanatory effect of plant production in the SEM).

Of the treatment variables, only flooding had a direct effect on elevation change (r = 0.27), that is, elevation gain was greatest under permanently flooded conditions. The other treatment manipulations did, however, have significant indirect effects on elevation change. Both CO2 and salinity influenced elevation change indirectly through impacts on S. americanus. In addition, flooding enhanced elevation change through three mechanisms: (i) a direct enhancement (r = 0.27), (ii) through the direct enhancement of S. americanus production (r = 0.45 × 0.65 = 0.29), and (iii) through indirect enhancement of S. americanus production via a negative impact on S. patens (r = –0.71 × –0.49 × 0.65 = 0.23). The total effect of flooding is the sum of these pathways (r = 0.27 + 0.29 + 0.23 = 0.79).


Predicted increases in atmospheric CO2 concentrations over the next 50 years are expected to influence mean sea-level due to global warming (IPCC 2007). Thus, an indirect impact of rising CO2 will be greater stress on coastal plants through increased flooding and intrusion of salt water. However, positive effects of CO2 enrichment on plant growth could offset negative effects of stressors associated with sea-level rise, especially if CO2 enrichment increases production of C3 species relative to C4 species because of a direct fertilization effect (Curtis et al. 1990; Arp et al. 1993; Lenssen et al. 1993; Erickson et al. 2007). In separate studies, both salinity and flooding have been shown to negatively affect the growth of certain species, including the C3 species S. americanus (McKee & Mendelssohn 1989; Broome et al. 1995; Rasse et al. 2005). Our study indicates that of the two dominants, S. patens is more salt-tolerant, whereas S. americanus is more flood-tolerant (within treatment levels) (Fig. 6). In the absence of a species that is tolerant of both salinity and flooding, another factor, such as CO2, may differentially affect stress tolerance of one species over another, leading to possible shifts in species composition.

Elevated CO2 can alleviate negative effects of salinity or water stress in certain species (Rozema et al. 1991; Lenssen et al. 1993; Rasse et al. 2005). Similarly, elevated CO2 ameliorated the negative impact of salinity on S. americanus in this study, while salinity did not affect the more salt-tolerant S. patens (Fig. 6). Elevated CO2 might ameliorate stress in plants by (i) enhancing net photosynthesis and reducing photorespiration or decreasing photoinhibition (Hymus et al., 2001), or (ii) by decreasing water loss via transpiration during photosynthesis (Robredo et al. 2007). Amelioration of salinity stress is further supported by a long-term field study by Erickson et al. (2007) in which elevated CO2 corresponded with increased C3 plant production relative to controls when precipitation was low and salinities were high (> 7‰).

Increased flooding favoured S. americanus production over that of S. patens, which is less flood-tolerant. Flooding combined with the CO2 × salinity interaction increased production of S. americanus versus S. patens in elevated CO2, even as salinity and flooding increased. Thus, the enhancement of C3 over C4 species with CO2 enrichment was not attributable to a direct CO2 fertilization response in this case, but to enhanced tolerance of hydro-edaphic stressors associated with sea-level rise.

Increases in plant production, as observed for S. americanus in this study, may contribute to elevation gain through direct and indirect effects. As the SEM indicates, elevation change was best predicted by the production of S. americanus, while production of the C4 species did not directly influence elevation. However, the C4 species had a negative effect on S. americanus production, which in turn affected biomass contributions to soil volume and elevation change. Increases in above-ground production and shoot densities have been shown to affect elevations directly by accelerating mineral sediment deposition and vertical accretion through baffling of sediment-laden water (Morris et al. 2002), and indirectly through biogenic accretion. In our experiment, however, mineral sedimentation was not a factor, and senescent biomass of both species was removed to prevent surface addition of litter (and to measure production). In the absence of surface deposition of mineral or organic material, below-ground biomass remained the sole means of biogenic accretion in this study. Rhizome and root production did not influence elevation change, which was surprising given that fine root accumulation was the primary control on soil building in mangrove systems (Middleton & McKee 2001; McKee et al. 2007). One explanation is that root and rhizome production was limited or underestimated. However, root accumulation measured in mesocosms (57–2818 g m−2 year−1) was comparable to that measured at field sites (110–1500 g m−2 year−1) (McKee & Cherry, unpublished data). An alternative explanation is that changes in shoot-base volume drove vertical soil displacement in this experiment (Fig. 7). A similar process has been documented in crop systems where shoot elongation vertically displaced soil aggregates, and the potential for soil displacement depended on shoot size (Whiteley & Dexter, 1982; 1984). Bases of individual S. americanus shoots were larger on average than those of S. patens (3.93 ± 0.15 and 2.02 ± 0.05 mm, respectively), suggesting that S. americanus would have a greater potential to affect elevation change than S. patens. In fact, shoot-base biomass was positively related to volume displacement for both S. americanus (R2 = 0.88, F1,58 = 408.88, P < 0.0001) and S. patens (R2 = 0.93, F1,58 = 763.90, P < 0.0001). Comparisons (based on culm densities × average shoot-base volume) showed that shoot bases of S. americanus contributed more to soil volume per mesocosm and elevation change than did S. patens. The positive relationship between elevation change and final volume of S. americanus shoot bases (Fig. 7), along with exclusion of root accumulation as a major factor in this experiment, suggests that the pathway in the SEM representing above-ground production of S. americanus was actually a proxy for shoot-base biomass and volume.

These findings suggest an additional explanation for correlations between stem density or shoot production of marsh species and elevation change. In field studies, the effect of plant shoots on soil accretion and elevation change has been attributed to a baffling effect on water movement, promoting mineral sedimentation (Morris et al. 2002). Because our experimental design eliminated baffling, our work suggests that increased stem densities may also indicate greater inputs of below-ground biomass to soil volume and vertical displacement. Thus, identification of the relationship between shoot-base expansion and soil elevation change provides further insight into how plant processes may influence vertical soil development and marsh capacity to accommodate sea-level rise. The dramatic difference between the two species in this regard may reflect differences in biomass allocation, with the C3 species allocating more biomass to below-ground shoot bases than the C4 species. If continued increases in atmospheric CO2 favour biomass production in brackish marshes, S. americanus, and possibly other C3 species, may play an increasingly important role in maintaining surface elevations in coastal marshes.

Changes in plant production, however, were not the only factors driving changes in elevation in this experiment. Elevation change was directly influenced by flooding, as waterlogged soils expanded under permanently flooded conditions. In field studies, hydrology has been shown to influence elevation change by causing swelling and shrinking of soils (e.g., Cahoon & Reed 1995; Whelan et al. 2005). In addition to physical swelling, the combination of increased salinity and flooding likely influenced elevation indirectly through effects on decomposition, a process not measured in this experiment. Oxidized conditions accelerate decomposition while anaerobic conditions in flooded soils slow decomposition. Flushing of soils during tidal drainage (mimicked in intermittent treatment) periodically introduces oxygen (Howes & Teal 1994), which would promote faster decomposition of soil organic matter. The direct positive path between flooding level and elevation change in the SEM may partly reflect slower decomposition under more waterlogged conditions. As oxygen disappears, alternate oxidants are used by facultative and obligate anaerobes in sequential descending order of energy (inline image, Mn4+, Fe3+, inline image) to break down organic matter, and soil redox potential declines (Gambrell & Patrick 1978). As salinity increases, however, anaerobic decomposition via sulphate reduction would also increase due to greater availability of sulphate in seawater (DeLaune et al. 2002), with a resultant decline in organic matter accumulation and elevation, as was observed at high salinities in this experiment (Table 2; Fig. 4). Thus, salinity and flooding may influence decomposition and organic matter contributions to soil volume. This hypothesis requires further testing, but these results suggest that elevation gain in this marsh system will be maximal where flooding and anaerobiosis limit decomposition but where salinity (and availability of sulphate) is low.

Given the complexity of interactions among biological and environmental factors, it is difficult to predict how coastal wetlands will respond to global drivers such as sea-level rise. The use of controlled mesocosm experiments that exclude some physical factors involved in elevation change and control others permits insight into biological processes and feedback relationships operating in these complex ecosystems. The use of SEM, in combination with anova, permitted testing of multivariate hypotheses regarding the relative importance of CO2, salinity and flooding for plant production and elevation change. This approach also provided much greater insight into abiotic and biotic processes than studies relying solely on anova significance tests, and revealed previously unrecognized mechanisms and interactions contributing to marsh elevation change.


Elevated CO2 ameliorated negative effects of salinity stress and enhanced production compared to ambient CO2 conditions, especially for the C3 species, S. americanus. Although S. americanus had a greater effect on elevation change, the presence of the C4 species, S. patens negatively affected S. americanus production and may limit the ability of S. americanus to contribute to positive elevation change. If increased CO2 ameliorates the effects of salinity stress in coastal marshes, thus increasing resilience of some species, then those species may achieve a greater competitive advantage over other species less responsive to changes in CO2. The greater contribution of S. americanus to elevation suggests that future shifts in composition of wetland plant communities could influence the capacity of marshes to maintain surface elevations relative to sea-level.

The strong relationship between production of S. americanus and elevation change identified by the SEM led to identification of shoot-base expansion as primary means of vertical soil displacement in this marsh system. While our findings do not eliminate the possibility of root contributions to elevation change, they do suggest that precise biological controls on elevation may vary among marsh types and component species, as well as with environmental conditions influencing biomass partitioning to above- and below-ground plant components.

Because the range of elevation change measured in the glasshouse corresponded to that observed in natural brackish marshes, this experimental design may be useful not only for studying biological contributions to elevation change, but also for understanding processes controlling elevation change in other sediment-deficient coastal marshes. Mechanisms revealed by SEM are currently being examined at a network of brackish marshes. Combination of these mesocosm and field studies will lead to more comprehensive models to identify key variables controlling marsh elevation, which can be used to generalize at regional or global scales and to forecast responses of coastal wetlands to global change. Such models will be essential for sound management and conservation strategies to prevent coastal wetland loss.


Authors thank Andrea Anteau, Heather Baldwin, Thomas McGinnis, Ellen Travis, and William Vervaeke for assistance with data collection and maintenance of experimental treatments in the Wetland Elevated CO2 Experimental Facility. Authors appreciate the comments from Glenn Guntenspergen, Jay Hodgson, and two anonymous referees who improved the quality of this manuscript. The US Geological Survey Biological Resources Discipline Global Change Program provided financial support for this study. Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the US Government.