Global change accelerates carbon assimilation by a wetland ecosystem engineer

The primary productivity of coastal wetlands is changing dramatically in response to rising atmospheric carbon dioxide (CO2) concentrations, nitrogen (N) enrichment, and invasions by novel species, potentially altering their ecosystem services and resilience to sea level rise. In order to determine how these interacting global change factors will affect coastal wetland productivity, we quantified growing-season carbon assimilation (≈gross primary productivity, or GPP) and carbon retained in living plant biomass (≈net primary productivity, or NPP) of North American mid-Atlantic saltmarshes invaded by Phragmites australis (common reed) under four treatment conditions: two levels of CO2 (ambient and +300 ppm) crossed with two levels of N (0 and 25 g N added m−2 yr−1). For GPP, we combined descriptions of canopy structure and leaf-level photosynthesis in a simulation model, using empirical data from an open-top chamber field study. Under ambient CO2 and low N loading (i.e., the Control), we determined GPP to be 1.66 ± 0.05 kg C m−2 yr−1 at a typical Phragmites stand density. Individually, elevated CO2 and N enrichment increased GPP by 44 and 60%, respectively. Changes under N enrichment came largely from stimulation to carbon assimilation early and late in the growing season, while changes from CO2 came from stimulation during the early and mid-growing season. In combination, elevated CO2 and N enrichment increased GPP by 95% over the Control, yielding 3.24 ± 0.08 kg C m−2 yr−1. We used biomass data to calculate NPP, and determined that it represented 44%–60% of GPP, with global change conditions decreasing carbon retention compared to the Control. Our results indicate that Phragmites invasions in eutrophied saltmarshes are driven, in part, by extended phenology yielding 3.1× greater NPP than native marsh. Further, we can expect elevated CO2 to amplify Phragmites productivity throughout the growing season, with potential implications including accelerated spread and greater carbon storage belowground.


Introduction
Global change is altering the fundamental ecological processes that control coastal wetland productivity, and is thereby altering ecosystem processes such as soil accretion, elevation gain, and carbon sequestration [1,2]. Global change affects primary productivity, for example, by altering photosynthetic rates, temporal patterns of growth, allocation to above-versus below-ground organs, and plant community composition [3]. Such changes have the potential to alter the resilience of tidal wetlands to sea level rise, with consequences for ecosystem services like providing wildlife habitat and protecting coasts from storm surges [4].
The productivity response of marsh ecosystems to global change is highly context specific, with the response to rising atmospheric carbon dioxide (CO 2 ) concentrations dependent on local abiotic conditions such as nitrogen (N) loading, salinity, and functional traits of the biotic community, including the dominant photosynthetic pathway [5][6][7]. Although research on native plant communities has provided strong insights into the relationships of rising CO 2 , N loading, and biotic change in coastal wetlands [6,8], biological invaders are re-engineering coastal ecosystems, thus modifying how global change factors influence ecosystem productivity, resilience to sea level rise, and provisioning of ecosystem services [9,10].
Invasive plant species and genotypes not only respond to environmental change but also amplify or fundamentally alter ecological processes through novel traits or feedbacks [11], and can therefore be considered a form of global change themselves [12]. Several highly successful plant invasions have proceeded through this form of ecosystem re-engineering; the most prominent case in North American coastal wetlands is that of Phragmites australis (hereafter Phragmites) or common reed. Phragmites is a clonal C 3 grass that grows in fresh to polyhaline wetlands throughout the world [13]. Although there are many genetic lineages of Phragmites globally, one haplotype (designated M) is highly invasive in North America, where it was introduced from Eurasia in the 1800s [14]. Several Phragmites lineages, including haplotype M, are known to have particularly strong trait responses to elevated CO 2 , temperature, and N conditions, such as increased light-saturated photosynthetic rates, relative growth rates, and stand densities [15][16][17]. However, experiments evaluating Phragmites traits and their responses to global change have been shortterm and have used young, containerized plants [15][16][17][18][19], thus limiting their ability to provide insight into ecosystem-level processes. Quantification of carbon fluxes and storage in Phragmites-dominated marshes, for example, requires data from mature clones growing in the field. Such quantification would be useful for carefully evaluating if and how Phragmites will alter rates of soil elevation gain under future environmental conditions. Some have suggested that the negative effects of Phragmites invasion be weighed against its potential to enable some coastal wetlands to keep pace with accelerating rates of sea level rise [20].
Relatively few data sets describe variation in carbon assimilation under expected future CO 2 conditions at timescales necessary to provide insight into the influence of intra-annual temporal patterns on annual productivity. However, growth and senescence phenology can strongly influence annual productivity, e.g., by moderating canopy structure [21,22] or the duration of the growing period [23][24][25]. Further, these phenological changes are influenced by global change factors such as CO 2 , N, and temperature [23,26,27]. Much of the research on plant productivity responses to global change is based on plant biomass measurements, with data collected one to a few times annually [6,28]. A lack of fine-scale, temporally explicit data on carbon assimilation rates under predicted future conditions limits our ability to understand and model the effects of global change on coastal wetlands. Although CO 2 , N, and other global change factors can be manipulated in the field, quantifying carbon fluxes at a high temporal frequency in concert with such experiments poses methodological challenges. Eddy covariance generates high frequency data and has been used successfully in coastal wetlands [29][30][31], but fumigation with CO 2 disrupts the CO 2 gradients that are the basis of the technique. Measurements of carbon assimilation can be made with flux chambers [32,33], but collecting data at a sufficient interval to capture variation in solar radiation is prohibitively difficult; this is especially true for large-stature species such as Phragmites (heights reach 4 m). Modeling can be used to overcome these challenges, though accurate modeling under global change conditions requires vegetative responses to these conditions to be determined and represented. At large spatial scales, this has been achieved by coupling models of carbon cycling and climate processes, and running these models under various emissions scenarios [34]. At the smaller scales relevant to determining the influence of plant community composition on productivity, modeling has enabled leaf-level carbon fluxes to be extended to the stand scale [24,35,36]. Empirical data from manipulative global change field experiments can provide the relationships and parameters needed to accurately represent differences in plant physiology and canopy structure in such models.
We present here a quantification of gross primary productivity (GPP) by Phragmites in a North American mid-Atlantic Coast wetland that is finely temporally scaled (hourly to daily) and was derived from a combination of experimentation and modeling. Using data and relationships derived from a new, manipulative field experiment on the Chesapeake Bay, we simulated Phragmites growth, canopy structure, and carbon assimilation under factorial combinations of atmospheric CO 2 and N loading. Although Phragmites stand dynamics have been modeled successfully several times before [37][38][39][40][41][42], these efforts did not consider the effects of global change on model parameters, and most focused on aboveground productivity. We specifically determined how global change will alter the magnitude of carbon assimilation by Phragmites tidal marshes at the annual timescale (GPP a ) as well as temporal patterns of carbon assimilation through the growing season. We also estimated annual net primary productivity (NPP a ) of Phragmites marshes, and used these values to constrain carbon use efficiency (CUE), i.e., the fraction of GPP a retained, under the global change treatments considered. Finally, we compared the stimulation effects of Phragmites marshes (based on GPP a and NPP a ) to each other and to those of native marshes (based on NPP a ). We expected Phragmites to increase GPP a and NPP a in response to both elevated CO 2 and N enrichment, but to experience the greatest productivity stimulation under the combined treatment. Further, we expected Phragmites to increase its CUE under all global change conditions and to exhibit stronger stimulation effects than the native community.

Field experiment
The field experiment was established in a brackish tidal marsh within the Smithsonian Global Change Research Wetland (Kirpatrick Marsh; 38.8742°N, 76.5474°W) in Edgewater, Maryland, USA. Salinity at the site varies from 4-15 ppt (mean=10 ppt) and the mean tidal range is 44 cm. The high-marsh platform is 40-60 cm above the mean low water level, and soils are predominantly organic (>80%) to 5 m depth. Mean daily air temperatures range from −4 to 31°C and mean annual precipitation is 108 cm. The native plant community is dominated by the C 3 sedge Schoenoplectus americanus (formerly Scirpus americanus and S. olneyi) and the C 4 grasses Spartina patens and Distichlis spicata. A single stand of Phragmites was documented in 1972 [43], whereas it now covers approximately 25% of the site.
Twelve open-top chambers (OTCs; 1.25×2.5×4.4 m) were installed at the leading edge of an expanding Phragmites stand in 2011. During the 2011-2013 growing seasons (May through October) half of the OTCs were fumigated with air approximately 300 ppm CO 2 above ambient levels (denoted eCO 2 ), while the remaining chambers were fumigated with unamended air. This CO 2 level is representative of those predicted for year 2100 under a moderate rise in global atmospheric CO 2 (e.g., scenario RCP6 [44]), and has been used in other experiments at the site [2,5]. Three OTCs of each CO 2 treatment type received 25 g N m −2 yr −1 (denoted N enr ), which was applied monthly during the growing season as dissolved NH 4 Cl (5 g N m −2 per month). This enrichment level represents moderate N loading in wetlands of the Chesapeake Bay [45] and has been used in an experiment in the native community at the site [2]. Treatments were randomly assigned to OTCs, thus preventing treatment effects from being confounded with spatially autocorrelated patterns, e.g., in Phragmites genetic relatedness or in microtopography.
Aluminum skirts at the base of each chamber (extending to 30 cm depth), in combination with low lateral movement of water at this site and the low mobility of NH 4 , ensured that minimal fertilizer was removed by tidal flooding [46]. Air temperatures measured in the middle and upper Phragmites canopy (175 and 230 cm, respectively) in one OTC from July-October were 2.0±2.8°C (24 h mean±SD) warmer than outside the chamber but above the native plant canopy (at 175 cm). Warming was greater during the day (3.6±2.8°C; 7:00-19:59 h) than at night (0.0±0.6°C; 20:00-6:59 h), and ∼1°C less in the lower canopy (85 cm). This magnitude of warming is comparable to that reported for another OTC experiment at the field site [27], and is likely due to heat being trapped by both the plant canopy and by the chambers. Phragmites abundance within OTCs increased from 2010-2013, but the composition of most chambers in 2013 was transitional between Phragmites-and native-dominated. An intensive data collection effort was undertaken during the 2013 growing season to define the properties and relationships used in a simulation model of canopy growth and carbon assimilation.

GPP simulation model
We determined canopy-level carbon assimilation by combining empirical data on plant growth, canopy structure, leaf senescence, and leaf-level photosynthesis in a simulation model. Methodological details of the model structure, the empirical data we collected, and the relationships we defined from these data are provided in supplement 1; a summary is provided in table 1. Briefly, we used the R computing environment (R Foundation for Statistical Computing, Vienna, Austria) to generate virtual, monotypic stands of Phragmites that had characteristics of plants growing in each of the four treatments at our field site (i.e., Control, eCO 2 , N enr , and eCO 2 +N enr ). Empirical data from the three replicate chambers of each treatment type were pooled to determine plant characteristics in simulations.
Stand densities were established at the outset of each simulation (table 1). At daily intervals, plants grew in height (figure 1) and added leaves according to relationships derived from field data (figures S1-S3). Daily representations of Phragmites canopy were divided into 10 cm thick layers such that total leaf area and associated aggregate characteristics (e.g., light attenuation) could be quantified. At hourly intervals spanning the 261 days of the growing season, empirical records of light availability and air temperature were applied to the canopy. Gross carbon assimilation by each canopy layer was determined as the sum of photosynthesis and respiration rates. These were based on monthly light response curves (figure S4) recorded at the leaf level near the top of the canopy. Measured rates were adjusted to account for declines with canopy position (figure S5) as well as diurnal changes in air temperature (table 1). Adjusted rates were summed across canopy layers and through all hours of the day to yield daily GPP values (GPP d ); these were then summed through the growing season to yield GPP a .
Phragmites stand densities vary widely in the field [16,47,48], and the associated changes in canopy leaf area alter carbon assimilation both positively via photosynthetic area and negatively via within-canopy shading. To assess how these countervailing processes would combine to determine carbon gain at the stand scale, we ran simulations at stand densities spanning the range typically seen in Atlantic Coast marshes [16,49]: 50, 75, 100, 125, 150 culms m −2 . Specifically, we ran simulations at 50, 75, 100, 125, and 150 culms m −2 . To characterize the influence of stochasticity associated with the selection of plant growth curves (figure 1), ten replicate model runs were carried out at each stand density. Where not stated otherwise, we report the means of replicate runs, with means computed at the daily scale. All computations were carried out in R 3.1.1.

Estimates of NPP and CUE
We calculated the NPP a of Phragmites marsh under each treatment condition from estimates of aboveground biomass (BM ag ). Biomass was derived from  morphometric measurements on individual plants that were collected during late July 2013, near peak standing biomass; we applied allometric relationships derived from plants collected outside of the chambers to determine BM ag . We then computed the mean mass per culm in each of the four treatments, and scaled these values to 100 culms m −2 (table S1 in supplement 2). We applied belowground mass fractions from an experiment with identical treatment levels as used in this study [55]. Note that plants in that experiment were containerized and grown from seed, potentially yielding smaller biomass allocation responses than mature plants exhibit. To convert biomass production to NPP a , we summed above and belowground biomass estimates and assumed that plant tissue was 45% carbon by mass [56]. We calculated Phragmites' CUE as the ratio of NPP a to GPP a .
We also compared stimulation effects between GPP a and NPP a for Phragmites-dominated marsh and with NPP a -based values for the native high-marsh community at the field site. Atmospheric CO 2 and N loading have been manipulated in factorial combination in the native marsh experiment (n=5 chambers of each of the four treatment types) using identical enrichment rates to those used in the Phragmites study, though it began in 2005 [6]. Dominant species in the native experiment included Schoenoplectus americanus, Spartina patens, and Distichlis spicata, all of which produce new aboveground organs annually. Chambers were censused near peak standing biomass (late July to early August) and BM ag was determined using allometric equations. Annual production of belowground biomass was measured with ingrowth bags (to 30 cm depth; collected in November). Given the low rates of decomposition in Chesapeake Bay tidal marshes [57] and the lack of dead fine roots in ingrowth bags (A Langley, personal observation), root turnover is likely to be negligible in this system. As with Phragmites, we assumed that biomass was 45% carbon.

Annual carbon assimilation
Our simulations of Phragmites australis stands growing under current atmospheric CO 2 concentrations and ambient N conditions (i.e., the Control) yielded a GPP a of 1.66±0.05 kg C m −2 yr −1 (mean±SD) at a typical stand density for Atlantic Coast marshes (100 culms m −2 ; figure 2). A 300 ppm rise in CO 2 (i.e., the eCO 2 treatment) induced a 44% stimulation to GPP a (we define stimulation as the percent increase over the Control), which equated to 2.39± 0.05 kg C m −2 yr −1 . Moderate N loading (i.e., the N enr treatment) yielded a 60% stimulation at ambient CO 2 , raising GPP a to 2.65±0.02 kg C m −2 yr −1 , while eCO 2 +N enr induced a 95% stimulation, raising GPP a to 3.24±0.08 kg C m −2 yr −1 . Phragmites culm heights, total leaf areas, and photosynthesis rates all increased under eCO 2 and N enr (figures 1, S3, and S4), contributing to these effects.
Altering stand density had a curvilinear effect on GPP a (figure 2) that was strongest under N enr and weakest under the Control. Specifically, a 50% increase in stand density (from 100 to 150 culms m −2 ) yielded increases in GPP a of 11%-31%, while a 50% decrease in density (100 to 50 culms m −2 ) yielded decreases in GPP a of 27%-39%. Despite the curvilinear relationship, treatment rankings were consistent across all densities. The margins by which eCO 2 exceeded the Control (0.56-0.73 kg C m −2 yr −1 ) and by which eCO 2 +N enr exceeded eCO 2 (0.79-0.88 kg C m −2 yr −1 ) were relatively consistent (CV=0.10 and 0.05, respectively). However, the margin by which N enr exceeded eCO 2 (0.36-0.11 kg C m −2 yr −1 ) decreased with stand density and was therefore more variable (CV=0.42).

Daily carbon assimilation
Temporal patterns of carbon assimilation differed among global change scenarios, with rates on a given day strongly influenced by PPFD (figures 3(a) and (c)). GPP d under all global change conditions exceeded rates under the Control by increasingly wide margins through most of the early growing season (March-May; figures 3(a) and (b)). Rates under N enr , with or without simultaneous CO 2 addition, were greater than rates under eCO 2 through this period, and surpassed 50% stimulation through most of April and May. By the end of May, cumulative carbon assimilation represented for 17%-19% of GPP a under the Control, eCO 2 , and eCO 2 +N enr treatments, but 22% of annual GPP a under N enr .
During the middle of the growing season (June-July), the temporal trend in GPP d was nearly linear under Control and eCO 2 conditions, though the rise was steeper under eCO 2 ( figure 3(a)). GPP d also increased under N enr , but the trend was more variable and less steep than for the Control; rates rarely exceeded those reached in late May. Given the greater rise in GPP d under eCO 2 than under N enr , the two treatments had similar rates during the peak period of July 15-20 (34.6±2.1 and 32.8±1.9 g C m −2 d −1 , respectively; mean±SD). Under eCO 2 +N enr , the temporal trend in GPP d paralleled that of the Control, but was 1.7× greater during the peak in mid-July (40.1±2.2 versus 23.0±1.3 g C m −2 d −1 ).
GPP d continued to exceed all other treatments under eCO 2 +N enr for the remainder of the growing season (August-November), exhibiting >100% stimulation for the majority of August and September. Carbon assimilated during the late growing season represented 27% of GPP a under eCO 2 +N enr . This delay in the end-of-season GPP d decline also occurred under N enr , though it was less pronounced; stimulation effects were half as strong as those under eCO 2 +N enr and carbon gains during the late growing season accounted for 22% of GPP a . In contrast, stimulation by eCO 2 declined throughout this period, with GPP d rates falling below those of the Control in late August. Carbon assimilated from August-November represented 18% of GPP a for eCO 2 and 21% for the Control.

Estimates of NPP and CUE
NPP a estimates for Phragmites-dominated coastal marshes ranged from 1.00-1.68 kg C m −2 yr −1 , with Control and eCO 2 +N enr conditions yielding the least and greatest rates, respectively (figure 4). NPP a varied among treatments primarily according to their N enrichment level; this largely derived from mean aboveground biomass being similar within each N level (table S1). NPP a was substantially greater for Phragmites than for the native saltmarsh community, with Phragmites 2.3 and 2.4× more productive than the native marsh under low N loading (i.e., Control and eCO 2 conditions, respectively), but 3.1 and 3.2× more productive under high N loading (N enr and eCO 2 +N enr , respectively). Carbon retained as NPP a represented approximately half of GPP a for Phragmites, although it lost less carbon to respiration under the Control (CUE=0.60±0.09; mean±SD, where SD was propagated from NPP a and GPP a ) than under any global change condition; CUE was 0.44±0.08 under eCO 2 , 0.57±0.09 under N enr , and 0.52±0.09 under CO 2 +N enr .
Elevated CO 2 stimulated NPP a in the Phragmites marsh far less than GPP a (6 versus 40%, respectively), but NPP a stimulation was similar to the multi-year average of the native marsh (9%; figure 4). N enr stimulated Phragmites GPP a and NPP a similarly (60 and 51%, respectively), whereas the native marsh experienced a far smaller stimulation to NPP a (13%). The combination of eCO 2 +N enr induced the greatest stimulation effects for Phragmites; effects were slightly less than additive for GPP a (95%) but more than additive for NPP a (68%). The native marsh experienced an NPP a stimulation under eCO 2 +N enr that was less than half of that experienced by Phragmites (32%).

Carbon assimilation
The annual carbon assimilation computed here for monotypic Phragmites stands indicates that the wetlands it invades will experience a sharp increase in GPP a under future CO 2 concentrations, especially where N loading is high. GPP a is typically <2 kg C m −2 yr −1 in temperate ecosystems, including wetlands [29,[58][59][60], but rates can exceed that value under eutrophied conditions (e.g., 2.3 kg C m −2 yr −1 [33]). At 3.24 kg C m −2 yr −1 , the rate we computed under eCO 2 +N enr for a stand containing 100 culms m −2 , Phragmites-dominated marshes would assimilate carbon more rapidly than many other temperate ecosystems, including those measured under elevated CO 2 [61][62][63].
Our simulations indicate that changes in Phragmites' growth and senescence phenology are key components of its enhanced productivity under elevated N conditions, and suggest that Phragmites will substantially increase carbon assimilation as atmospheric CO 2 rises. Most notably, we observed far greater GPP d early and late in the growing season under high versus low levels of N loading. This temporal pattern can be attributed to a combination of earlier initiation of canopy growth, prolonged canopy expansion, and maintenance of high leaf-level photosynthetic rates from June through September (figures 1, S3, S4). In addition, Phragmites carbon assimilation declined late in the growing season under eCO 2 . This response contrasts with that of the dominant native C 3 species at our site (Schenoplectus americanus), whose biomass is elevated through delays in senescence under eCO 2 [27].

Carbon retention
Under future atmospheric CO 2 concentrations, our results indicate that Phragmites-dominated wetlands are likely to experience stronger stimulation of NPP a than most other wetland or upland systems. Many CO 2 enrichment studies have found small or negligible stimulation effects on productivity, especially when C 4 grasses are dominant [28,64]. Few field studies have quantified combined eCO 2 +N enr stimulation effects; the three we found for grass-dominated communities that did [65][66][67] all based their values on current-year biomass production (≈NPP a ). Stimulation effects ranged from 7 to 42% in the three studies. Similarly, the CO 2 ×N experiment of the native community at the Global Change Research Wetland experienced a mean eCO 2 +N enr stimulation of 32% over eight years. All of these NPP a stimulation estimates are therefore well below the 68% that we found for Phragmites.
Our estimates of Phragmites' growing season CUE (0.44-0.60) are in the middle of the range reported for wetland plant species (0.34-0.77) [33,60,68], suggesting that Phragmites' high NPP a and strong stimulation by global change factors are due to advantages in carbon assimilation rather than an ability to minimize carbon losses via respiration. The decrease in CUE under eCO 2 relative to the Control is consistent with patterns seen in some, but not all, other studies [63]. Such decreases can arise if elevated CO 2 facilitates greater nonstructural carbohydrate production, as their metabolism can induce a greater rise in rates of respiration than photosynthesis [68]. The fertilization-induced increase in CUE seen under elevated, but not ambient, CO 2 may have been due to decreased allocation to roots causing respiration to increase less than photosynthesis [63].

Model evaluation
Results from two other studies of carbon assimilation by Phragmites suggest that our GPP a values are reasonable. A study using eddy covariance reported GPP a for a Phragmites-dominated marsh in northeastern China as 0.71 kg m −2 yr −1 [29]. Given that this haplotype had approximately half the maximum height (1.5 m; stand density not reported) of the Phragmites at our field site under ambient conditions, a GPP a that is 52% of our modeled values for Control conditions over the range of likely stand densities (50-100 culms m −2 ) suggests good agreement. Further, Stefanik and Mitsch [33] measured GPP for North American introduced Phragmites using flux chambers in a constructed freshwater wetland. Although the stand density was lower at their site than our simulation evaluated (42 culms m −2 ; K Stefanik, pers comm), N loading was substantially greater (∼100 g N m −2 yr −2 [69]). Nevertheless, the estimate of GPP a in that study (3.09 kg C m −2 yr −1 ) was similar to the value generated by our N enr simulations at 150 culms m −2 (2.93 kg C m −2 yr −1 ). The results of these studies demonstrate that our quantification of GPP a is in line with values determined via more direct measurements than we used, at least under the treatment conditions considered.
Our quantification of GPP in Phragmites-dominated marshes had several inherent limitations. For one, we could not determine the effects of salinity, which changes inter-annually due to variation in rainfall and sea level, and reduces productivity and CO 2 stimulation effects when high [7]. Because our focal growing season had moderate salinity conditions (monthly means ranged from 8.2-12.4 ppt), years with more (or less) rainfall than occurred in 2013 would likely yield greater (or lesser) GPP a than reported here. However, we would not expect the effect of salinity to be consistent among treatments, as elevated CO 2 is known to reduce the effects of salinity on Phragmites growth [70]. Second, we assumed that variables were unaffected by CO 2 and N conditions when we did not have empirical data demonstrating otherwise; these included leaf lifespan, Q 10 values, and the decay constant for PPFD attenuation. This assumption is unlikely to hold true in many cases. For example, N enrichment can increase leaf longevity in Phragmites [71]. Third, although chamber temperatures reflected warming in the range expected by year 2100 [72], our temperature time series reflected contemporary conditions. If acclimation substantially altered the nature of Phragmites' photosynthesis response to varying temperature, our results could have been biased. However, it is difficult to predict the magnitude or direction of this potential bias, especially given that some of Phragmites' physiological responses to temperature are CO 2 dependent [18]. We expect to be able to address many of these limitations with empirical data once Phragmites densities increase in our field experiment.

Conclusions
The data-driven simulation of Phragmites canopy growth used here allowed us to translate leaf-level photosynthesis data to the stand scale, and thereby calculate GPP at a fine temporal resolution. The approach provided information on GPP that was independent of NPP measurements, making it possible to constrain CUE under global change conditions. Although the method's data requirements are not small, similar approaches could be used in other annual systems where plant canopy structure is relatively simple. Perhaps the greatest advantage of the approach is that it can provide insight into intraannual temporal shifts in carbon assimilation induced by global change.
By quantifying GPP at a daily resolution, we were able to determine that N enrichment sharply increased carbon assimilation both early and late in the growing season, whereas CO 2 elevation increased assimilation more moderately and through the early and midgrowing season. Given that N loading is already elevated in many North American estuaries [73], productivity advantages due to extended leaf phenology help to explain the close landscape-level association between the distribution of Phragmites and N enrichment [74,75]. Our results also confirm earlier speculation that delayed leaf senescence by introduced versus native Phragmites in North America contributes substantially to its productivity, especially in eutrophied wetlands [56]. Further, the strong increase in late season carbon assimilation that we found under eCO 2 +N enr compared to N enr conditions suggests that late season carbon gains will comprise an increasing fraction of Phragmites' annual production.
Our results indicate that Phragmites could increase in productivity through the coming century, given that atmospheric CO 2 levels will likely reach 700 ppm and that eutrophication will likely become increasingly widespread in that time [44,76]. Although the productivity of native saltmarsh plants is also projected to increase in response to global change, taxa investigated previously have responded much less strongly than Phragmites to combined increases of CO 2 and N [6]. The plastic response to eCO 2 +N enr found here indicates that Phragmites will be able to capitalize on these two global change factors simultaneously, likely yielding stronger competitiveness than it exhibits currently. Together with other advantages that Phragmites gains under global change conditions, like access to deepsoil nutrients [55] and elevated patch-level genetic diversity, floret production, and therefore potential for spread via seeds [77,78], this productivity advantage will likely translate into accelerated rates of Phragmites invasion in tidal marshes of the North American Atlantic Coast.
Global change-induced increases in Phragmites productivity may have particularly strong effects belowground. Phragmites allocates a substantial proportion of its growth to roots and rhizomes (82% of standing biomass in the Meadowlands of New Jersey) [79], and these penetrate more deeply into soils than do belowground organs of native plants [80,81]. In addition to the greater root and rhizome biomass that can be expected to accompany high Phragmites GPP in the coming decades, belowground biomass is also likely to deepen further into the soil profile [55]. Rhizome construction costs are also lower under eCO 2 +N enr than under Control conditions [19], suggesting that rhizome growth could be particularly strongly enhanced under future conditions, together with rhizome-dependent processes like convective gas flow [82]. A key effect of greater rhizome and root production may be accelerated mineralization rates of sequestered nutrients, which could provide an additional N source for Phragmites and thereby accelerate its vegetative growth, seed production, and spread, even in sites that have low N inputs [55]. Further, exports of dissolved carbon (DOC and DIC) from wetlands invaded by Phragmites may increase through greater belowground productivity, potentially affecting the carbon balance of coastal aquatic communities [83].
Increases in Phragmites productivity also have the potential to influence carbon sequestration belowground, which determines the soil building capacity (i.e., surface accretion) of tidal wetlands, especially in peat-based systems [1]. Given rising sea levels, our estimates of NPP suggest that Phragmites marshes will have an increased likelihood of outpacing sea level rise than native communities [20,84], although more refined quantifications of accretion will be needed to address this possibility with greater certainty. Coastal wetland conservation may therefore require strategic planning in order to decide whether to prevent or allow re-engineering by Phragmites at small spatial scales, such that ecosystem functions can be optimized at broad spatial scales. Development and refinement of predictive models like the one used here will be required to accurately forecast the local-and landscape-scale effects of global change on coastal wetland ecosystems, and to support management decisions that will mitigate the threat that accelerated sea level rise poses to their future stability.