Poplar Island is located in mid-Chesapeake Bay, 3.2 km northwest of Tilghman in Talbot County, Maryland, United States (Figure 1). The site experiences a mean diurnal tide range of 0.47 m (NOAA Tides and Currents, station ID 8572271), and a mean surface salinity range (1985–2018) of 10.1–12.5 (US EPA CBP 2019). The study was conducted primarily in one marsh subunit (cell) on the site, Cell 1B, a 12.41 ha tidal marsh completed in 2012. The marsh platform was designed as 80% low marsh (LM, 6.92 ha) and 20% high marsh (HM, 2.51 ha), planted with S. alterniflora and S. patens, respectively, and 1.79 ha of tidal creeks. The cell also includes a 1.19 ha upland habitat island, which was excluded from the marsh mass balance estimates. The marsh substrate (to a depth of approximately 2 m) is fine-grained dredged material resulting from maintenance dredging of the navigation channels in upper Chesapeake Bay (Cornwell and Owens, 2011), providing a high fertility substrate for the restored tidal marshes (Cornwell et al., 2020). Each marsh is surrounded by sand dikes extending approximately 3 m above mean sea level, and tidal exchange occurs through side-by-side 1.83 m² concrete box culverts, simplifying flow estimates and making the site well suited to a tidal flux study. The marsh is connected to adjacent marshes (Cells 1A and 1C; Figure 1) via shallow lateral channels designed to facilitate fish passage, but exchange through these channels is less than 5% of tidal exchange at the inlet (Fleri et al., 2019). Upland drainage into the marsh is limited to the area of the surrounding dikes, and the consolidated dredged material substrate limits transport within the sediment profile primarily to diffusion.

A Si mass balance developed for the Cell 1B marsh (11.22 ha, excluding the upland habitat island) identified potential inputs as atmospheric deposition, tidal exchange with the estuary, and upward diffusion from the sediment below the root zone. Potential outputs were tidal exchange and burial. Measurements of site-specific atmospheric deposition were beyond the scope of our study, so an estimate based on literature values was used. Upward diffusion of DSi from the sediment below the root zone was not estimated and internal cycling, while important, was not quantified in this study with the exception of remineralization from macrophytic vegetation. All other pools and fluxes were based on Si analysis of samples collected as part of the long-term monitoring program at Poplar Island, or as part of targeted experiments from 2012 to 2014.

The pool of Si in marsh vegetation was estimated from 2014 end-of-season biomass measurements (Staver et al., 2020), combined with mean end-of-season tissue SiO₂ concentrations from live S. alterniflora shoots (2.37%) and roots/rhizomes (1.06%) from the control subplots in the Si amendment experiment (described below), and standing dead material collected for a decomposition experiment (2.00%) in 2012–2013 (Supplementary Table S1). Tissue SiO₂ concentrations were not determined for S. patens, but Giblin et al. (2021) reported a range of 0.29–2.19% SiO₂, similar to that for S. alterniflora, so the same tissue concentrations were used to estimate the high marsh vegetation pool. Annual shoot and root production in Cell 1B were estimated from clipped plots (0.25 m²) and sediment cores (42 cm² × 20 cm) harvested in October 2014 (Staver et al., 2020, 2021).

Porewater and solid phase soil concentrations were combined to estimate the total sediment Si pool within the root zone (15 cm depth). Porewater samples were obtained using porewater equilibrators (Hesslein, 1976), deployed and retrieved following Cornwell et al. (2020). Dissolved Si concentrations were determined as SiO₂ using the molybdate blue methodology (Strickland and Parsons, 1968), modified for a Technicon AutoAnalyzer II (Technicon Industrial Method 186-72 W-Modified), with ascorbic acid as the reductant. The sediment porewater DSi pool was estimated from concentrations measured in the control subplots of the Si amendment experiment described below, and assumed a pore space volume of 50%. A representative value from these profiles of 200 μmol L⁻¹ was used to calculate the root zone DSi pool. Sediment solid phase concentrations were determined as BSi for 18 solid phase samples collected in 2011, prior to planting the marsh, using the base digestion technique (Demaster, 1981) modified by Saccone et al. (2006) followed by silica (SiO₂) analysis. The BSi range among those samples was 2.34–3.45% dry weight. The mean value, 2.96% (± 0.07 se, n = 18), and a bulk density of 1.1 g cm⁻³ (Staver et al., 2021) was used to calculate the root zone BSi pool.

An average value for the Si pool in tidal creeks (68.28 μmol L⁻¹) was estimated by using the mean annual ebb concentrations of DSi and BSi (41.96 and 26.46 μmol L⁻¹, respectively) in the tidal exchange study, combined with the tidal creek area (1.79 ha) and an average depth of 1 m.

The Si pool in edaphic algae was not quantified here, although it is recognized that benthic diatoms are often present on the marsh surface and are likely important in Si cycling within the marsh seasonally (Ragueneau et al., 2006). Staver et al. (2020) reported seasonally high rates of carbon fixation in unvegetated sediment cores, but respiration always exceeded carbon (C) fixation. Thus, while the benthic algal community is active, fixed carbon is rapidly recycled and stocks are low. In our study, although it is recognized that diatoms are a part of the benthic algal community and may be important in Si processing, it is assumed that, like C, the Si pool is comparatively small and recycles rapidly.

Annual tidal fluxes of DSi and BSi were estimated from seasonal measurements made during 2014. In addition, dissolved inorganic N (DIN), soluble reactive phosphorus (SRP), total suspended solids (TSS) and chlorophyll a (Chl a) concentrations were measured to allow examination of correlations with particulates and the stoichiometry of Si and essential plant nutrients N and P. A full description of tidal water velocity measurements and constituent flux calculations can be found in Staver et al. (2020). In brief, seasonal (February, May, July, and November) tidal water velocity measurements were made with a Nortek Aquadopp acoustic Doppler profiler (ADP), mounted on a weighted PVC plate and placed in the center of one culvert, equidistant from the open ends and sidewalls.

Concurrent with the water velocity measurements, an ISCO automatic water sampler was deployed for a 24-h period (two full tidal cycles) in the same culvert at the mouth of Cell 1B. Discreet water samples were collected at hourly intervals from 30 to 40 cm off the bottom of the culvert. The samples were chilled with ice in the ISCO sampler, returned to the lab at the conclusion of the sample period and kept refrigerated during processing. Samples were subsampled for DSi within 24 h, usually immediately upon return from the field (within 2 h). For BSi analysis, 30 ml subsamples were filtered through 0.45 μm polycarbonate filters, which were analyzed according to Saccone et al. (2006). For validation, the DSi concentrations from the tidal flux study were compared with ambient concentrations measured monthly in 2014 at three long term water quality monitoring stations at Poplar Island. Two stations were located in tidal creeks within the interior of Poplar Island marshes Cell 4D (sand substrate, completed 2003) and Cell 1C (fine-grained dredged material, completed 2005), and a third was located just offshore in Poplar Harbor (WQR1; Figure 1).

Burial rates were determined in a litter bag study described in detail in Staver et al. (2020). In brief, litter bags of S. alterniflora aboveground (AG) biomass were deployed in the low marsh in three fine-grained dredged material marshes spanning a range of age, Cells 3D, 1A and 1C (constructed in 2005, 2009 and 2011, respectively), and Cell 4D, the low nutrient sand substrate marsh (constructed in 2003) from February 2012 to December 2013. Decomposition rates were not determined separately for S. patens or for the high marsh; so for the annual burial estimate the low marsh, the S. alterniflora rate was also applied to the high marsh. For this study, litter bag contents were analyzed for BSi as described above (Saccone et al., 2006). Based on the results of this study, which showed minimal dry mass loss during the second year, it was assumed that a steady state in decomposition is reached after 2 years (Staver et al., 2020). This is likely not the case in the high marsh, which constitutes just 20% of the marsh platform, potentially introducing a small error in the marsh estimate. The burial rate was estimated as B = A P ∗ % D M ∗ % S i O 2 where AP is the area-weighted mean annual biomass production of the high and low marsh (2,131 g dry weight; Supplementary Table S1); % DM is the mean percent of original dry mass of S. alterniflora litter remaining in litter bags; and % SiO₂ is the SiO₂ concentration in litter at the conclusion of the decomposition study in the fine-grained dredged material cells (Cells 1A, 1C, 3D).

In a review of global atmospheric Si transport, Tegen and Kohfeld (2006 in The Silicon Cycle) estimate that the mid-Atlantic coast receives between 0.1–1.0 g Si m⁻² y⁻¹. For atmospheric deposition, we used the estimate of Anderson and Downing (2006) for Iowa, where agriculture is a dominant land use, similar to much of the region surrounding our study site. Atmospheric deposition averaged 0.61 g Si m⁻² y⁻¹ (equivalent to 1.30 g SiO₂ m⁻² y⁻¹), with highest deposition rates occurring in spring when agricultural activity causes dust to become airborne. While agricultural activity is absent on Poplar Island, it was assumed that dust produced by the use of heavy equipment as part of on-going dredged material management activities is analogous to that produced by agricultural activity, meriting a mid to upper level estimate within the range of Tegen and Kohfeld (2006).

Pearson Product Moment Correlations were performed to determine relationships between tidal flux components. For the phytotron experiment, statistical analyses included Kruskal–Wallis one-way analysis of variance on ranks (ANOVA) to test for significant effects of a soil amendment on shoot growth and root growth. One-way ANOVA was performed on tissue SiO₂ concentrations, followed by pairwise comparisons based on rank sums (Tukey test). For the field experiment, significant differences in soil porewater DSi profiles between Control and Si amended plots were tested with a split-plot ANOVA, with plot, treatment, and depth as variables. Differences between amended and control treatments in tissue SiO₂ concentrations throughout the growing season in each year (2012 and 2013) were tested with a split-plot ANOVA, with plot, treatment and date as variables. Statistical analyses for the phytotron, tidal flux and fungal (field experiment) data were performed in SigmaPlot 13.0 at p = 0.05; analyses for the soil porewater and S. alterniflora tissue SiO₂ concentrations (field experiment) were performed in R using the Agricolae package for split-plot experimental designs.

Silicon cycling in the created tidal marsh (Cell 1B, Poplar Island) in Chesapeake Bay is notable for the large net annual tidal export of combined dissolved and biogenic SiO₂. The export of DSi is consistent with Si fluxes reported for other marshes (Struyf et al., 2006; Jacobs et al., 2008; Vieillard et al., 2011). Unlike other marshes, which import BSi with TSS, especially in summer (Struyf et al., 2007), there was a constant export of BSi from Cell 1B. The estimated annual uptake of DSi by macrophytes (4,034 kg) is of similar magnitude as the estimated net annual export of DSi from the marsh (5,468 kg), and is about half the total tidal export (DSi + BSi, 9,448 kg). Combined with the estimate that approximately 6% of the SiO₂ mobilized from the sediment through macrophyte uptake is re-buried with organic matter, the mass balance suggests there is a significant flux of DSi out of the sediment in addition to macrophyte uptake.

The dredged material substrate is the primary source of N for macrophytic growth and the driver of net N export in the Cell 1B marsh (Staver et al., 2021). Similarly, it appears to be the primary source of SiO₂ for uptake by macrophytic vegetation and both benthic and pelagic diatoms, as well as direct fluxes to the water column during flood tides. While direct measurements of sediment DSi fluxes were not made as part of this study, sediment porewater profiles show depletion in the root zone (≤ 15 cm depth) resulting from plant uptake and, likely, direct fluxes when the marsh is flooded, since porewater DSi concentrations greatly exceed those in overlying water. Concentrations below the root zone remain among the highest reported (Müller et al., 2013), 200–1,000 μmol L⁻¹ in 2012, exceeding concentrations reported for a natural marsh on Plum Island, MA, 0–300 μmol L⁻¹ (Vieillard et al., 2011; Carey and Fulweiler, 2013, 2014a), and a Delaware marsh, 100–900 μmol L⁻¹ (Scudlark and Church, 1989). The high concentrations of both BSi and DSi in the dredged sediment likely result from the accumulation of fine-grained particles of terrestrial origin combined with diatomaceous BSi in the deep navigation channels. As the dredged material ages, concentrations will likely decline due to export. There was little change over the course of this study (Figure 9), however, suggesting the large volume of dredged material beneath the root zone will provide a long-term source.

Potential areas of uncertainty in the Si mass balance include heterogeneity in macrophyte biomass production (Wehrhan et al., 2021), and the estimate of tidal fluxes. Biomass is monitored annually in the Poplar Island marshes, and the 2014 data used here is consistent with average production in these marshes. In some years, however, sudden vegetation dieback (Elmer et al., 2012) dramatically reduces S. alterniflora production and likely affects Si uptake, recycling and exchange with the estuary. In addition to the influence of macrophyte production, estimates of tidal exchange are constrained by limited sampling. Hourly water sample collection over two tidal cycles was used to estimate seasonal fluxes, and storm events were not represented in our study. The ADPs were deployed for a minimum of 1 week in each season, to ensure the tidal prisms during water collection were representative of that period, but extrapolating to a full three-month season undoubtedly introduces some error. This could be constrained in future efforts through more frequent sampling, e.g., monthly, combined with storm sampling to characterize fluxes associated with extraordinary tides. A smaller error is likely due to an overestimate of burial, for which low marsh decomposition rates were applied to the high marsh where decomposition can be slower due to differences in vegetation and less frequent tidal inundation (Frasco and Good, 1982). Since only 20% of the marsh platform is high marsh, a more accurate burial estimate would have a minimal impact on the mass balance.

In our study, the peak export of DSi and BSi occurred in July when soil temperatures and sediment flux rates are reported to be highest (Scudlark and Church, 1989). However, SiO₂ loss from S. alterniflora detritus was most rapid during the spring (Figure 3) prior to peak DSi and BSi export from the marsh (Table 3). The delay may be due to assimilation by benthic diatoms in spring, followed by release as diatoms are shaded by macrophytes or replaced by other algal species during the warmer months.

In Danish marshes, the summer export of DSi is considered a valuable DSi source to coastal waters when ambient DSi concentrations are at a minimum (Struyf et al., 2006; Jacobs et al., 2008). In mid-Chesapeake Bay, the annual minimum occurs in early summer (Conley and Malone, 1992), prior to the peak export by the Poplar Island marshes (Table 3 and Figure 5). Rather, it coincides with a summer peak in ambient DSi resulting from remineralization of diatom BSi following the spring diatom bloom (Conley and Malone, 1992). In the tidal flux study presented here, the positive relationships between BSi and TSS, and BSi and Chl a (Figure 4) suggest that much of the BSi export from the marsh is in the form of pelagic diatoms. If this is the case, it suggests that diatoms remain abundant in the marsh creeks despite the high inorganic N concentrations due to the availability of DSi. The Si:N ratio in diatoms is ~ 1 (Sterner and Elser, 2002), while the molar Si:N ratio of exports from Cell 1B is approximately 6.6 for dissolved forms of Si and N, suggesting that overall the marsh is exporting a surplus of Si relative to N and diatom production in marsh channels is not Si limited. Together, these observations support the hypothesis that DSi depletion in eutrophic estuaries is responsible for the shift in species composition toward non-diatomaceous species, including those which produce harmful algal blooms (Giblin et al., 2021), and may have implications for higher trophic levels. A characterization of temporal changes in algal species composition in the Poplar Island marsh creeks could resolve this question.

The retention of macrophyte detritus within Cell 1B due to the limited tidal exchange appears to contribute to robust rates of vertical accretion (Staver et al., 2020), although it is not clear what effect it may have on Si burial rates. Struyf et al. (2007) report that young Dutch marshes have higher deposition and burial rates than mature marshes, with rates of both declining after the marshes reach an elevation in equilibrium with sea level. The resulting recycling efficiency ranges from 15% in young marshes (85% buried) to 60% in mature marshes (40% buried). The young marshes at Poplar Island also have higher rates of vertical accretion and may follow a similar trajectory. However, upward diffusion of sediment N and DSi from below the root zone likely persists for years to decades and may result in persistently higher recycling rates and lower burial rates for a longer period than in the Dutch marshes.

Although the marshes studied by Struyf et al. (2007) were formed on former agricultural land, they did not consider the effects of soil N on Si recycling or burial. This was examined in a study of two Rhode Island marshes exposed to different levels of N (Carey and Fulweiler, 2013). Nitrogen enrichment was associated with higher sediment BSi and DSi concentrations, as well as higher plant tissue concentrations. Increased Si availability due to high TSS deposition and high rates of recycling induced by N enrichment were thought responsible for the elevated sediment and plant concentrations, although there was likely also a difference in Si inputs. The results presented here suggest that at the ecosystem scale, increased primary production resulting from N enrichment enhances sediment SiO₂ mobilization, remineralization and export. Thus, through continuous export of DSi these marshes may help mitigate the negative effects of eutrophication on spring and summer diatom populations in the surrounding estuary, potentially enhancing their role in support of fisheries (Baker et al., 2020).

At the plant scale, there has long been speculation that S. alterniflora could develop a growth induced Si deficiency in response to N enrichment (Lanning and Eleuterius, 1981, 1983), and several studies have examined the interactions of marsh macrophytes, stress (biotic and abiotic) and Si availability. Results of amendment studies have been mixed, with some species showing enhanced tissue concentrations in response to Si enrichment (Schaller et al., 2012; Mateos-Naranjo et al., 2013), while others did not (de Bakker et al., 1999). The role of tissue Si concentrations in regulating S. alterniflora growth and resistance to pathogens, grazing and abiotic stressors also remains unclear (Querne et al., 2012; Bazzano and Elmer, 2017).

In our study, it was hypothesized that N availability in the Poplar Island marshes, combined with the numerous indications of N stress in S. alterniflora, made this an ideal site to examine whether growth induced Si deficiency can be mitigated by soil Si amendments, as in some crop species. The results of our small-scale phytotron experiments growing S. alterniflora in dredged material enriched with calcium meta-silicate were, however, inconclusive. While Si amendments had a pronounced effect on soil pH, there was not a large difference in plant growth or tissue Si concentrations with increasing Si amendment. Plants grown in sand were stunted compared with those grown in fine-grained dredged material, likely due to N deficiency, but they contained similar tissue Si concentrations (Figure 9). The experiment was only 12 weeks long, and since DSi is taken up passively by S. alterniflora, this experiment may not have reflected the results that would be obtained in a longer growing season. However, these data highlight the effect of N enrichment on soil Si mobilization by macrophytic vegetation through increased biomass production. This likely contributes to an increase in DSi availability within the marsh and to the large net total SiO₂ export from the Cell 1B marsh.

In the two-year field experiment, tissue concentrations were similar in Si-amended and control plots early in each growing season, but differences developed and increased as the season progressed. In contrast to the phytotron experiment, it appeared that soil SiO₂ amendment can help mitigate root zone depletion and produce a modest effect on tissue concentrations. This effect persisted through year two following soil amendment. Thus, it appears that in an N-rich environment, SiO₂ uptake potential can exceed supply even where availability is high.

Also noteworthy in the field experiment was that tissue Si concentrations of S. alterniflora plants from Cell 4D, which has a sand (low nutrient) substrate, were substantially lower than plants from the dredged material plots, while plants from the HPL reference marsh, with higher soil porewater NH₄⁺ concentrations (unpublished) were closer to the dredged material marsh controls. These results align with the results of Carey and Fulweiler (2013) that plants exposed to higher N availability contain higher tissue Si concentrations. What is not clear is whether this is due to differences in N availability and plant growth, or differences in Si availability or uptake. Nitrogen enrichment delays senescence, and passive Si uptake persists as long as the plants are transpiring. A prolonged growing period for N enriched plants could result in higher tissue Si concentrations at the end of the growing season through passive uptake in the absence of increasing biomass. Active uptake has also been reported for S. alterniflora (Carey and Fulweiler, 2014b), but the steady increase in tissue SiO₂ throughout the growing season in the field experiment suggest that passive uptake is the primary mode of Si acquisition in the Poplar Island marsh.

The question of whether increased tissue concentrations confer resistance to stress in S. alterniflora was not resolved by these studies. Neither lodging nor grazing by O. zibethicus were observed in or near the experimental plots during the course of the field experiment. Fungal analysis suggested that there could be a protective mechanism conveyed by Si enrichment, at least with respect to the more pathogenic strains found in S. alterniflora. Bazzano and Elmer (2017) found Si amended S. alterniflora in greenhouse trials did not, however, differ in susceptibility to F. palustre or in herbivory by marsh crabs when compared to untreated plots.

Although some increases in S. alterniflora tissue Si concentrations were observed in response to amendments in this study, associated improvements in fitness and production were not observed even under extreme N enrichment, as has been reported for O. sativa and some other crop species. Both species are members of the same family, Poacea, known to have higher tissue concentrations than most other vascular plants. S. alterniflora, however, has been designated an intermediate Si user accumulating Si primarily through passive uptake, with tissue concentrations of 0.5–1.0% dry weight (Querne et al., 2012), whereas O. sativa is considered an accumulator, showing both passive and active uptake and tissue concentrations > 1.0% (Ma and Takahashi, 2002). Notably, tissue concentrations in our study were within the range considered to be characteristic of accumulators, but we hypothesize that this results from N-enrichment, which can delay senescence, allowing Si uptake to continue later in the growing season (Marschner, 2012). The reasons for generally lower tissue Si concentrations in S. alterniflora compared with O. sativa are unclear, but may be related to Si availability in brackish to saltwater versus freshwater environments, or to physiological requirements related to perennial (S. alterniflora) versus annual (O. sativa) life cycles.

The present study has demonstrated that newly constructed fine-grained dredged material marshes can provide a source of both dissolved and biogenic Si to local waters in stoichiometric excess (relative to algal requirements) of exported N and P, an overlooked ecosystem service. High nutrient availability drives high rates of macrophyte production, which appear to enhance mobilization of sediment DSi and remineralization. The tidal creeks also appear to be important in the transformation and export of Si. Effects on the composition and shifts of the local benthic and pelagic algal community are unknown, but may be important locally for higher trophic levels, a potential area for future study.

Soil amendment with calcium meta-silicate resulted in slightly enhanced tissue concentrations in S. alterniflora, but improved plant fitness was not detected in this study, even in this N rich environment where symptoms of N stress were abundant. The accumulation of evidence from this and other studies appears to cast doubt on the concerns of Lanning and Eleuterius (1981, 1983) that N stress in S. alterniflora results from a Si deficiency.