Vegetation and proximity to the river control amorphous silica storage in a riparian wetland (Biebrza National Park, Poland)

Wetlands can modify and control nutrient fluxes between terrestrial and aquatic ecosystems, yet little is known of their potential as biological buffers and sinks in the biogeochemical silica cycle. We investigated the storage of amorphous silica (ASi) in a central-European riparian wetland. The variation in storage of ASi in the soil of an undisturbed wetland was significantly controlled by two factors: dominance of sedges and grasses and distance to the river (combinedR2=78%). Highest ASi storage was found near the river and in sites with a dominance of grasses and sedges, plants which are well known to accumulate ASi. The management practice of mowing reduced the amount of variation attributed to both factors ( R2=51%). Although ASi concentrations in soils were low (between 0.1 and 1% of soil dry weight), ASi controlled the availability of dissolved silica (DSi) in the porewater, and thus potentially the exchange of DSi with the nearby river system through both diffusive and advective fluxes. A depth gradient in ASi concentrations, with lower ASi in the deeper layers, indicates dissolution. Our results show that storage and recycling of ASi in wetland ecosystems can differ significantly on small spatial scales. Human management interferes with the natural control mechanisms. Our study demonstrates that wetlands have the potential to modify the fluxes of both DSi and ASi along the land-ocean continuum and supports the hypothesis that wetlands are important ecosystems in the biogeochemical cycling of silica. Correspondence to: E. Struyf (eric.struyf@ua.ac.be)


Introduction
Biologically controlled uptake and storage of silica and its recycling in terrestrial ecosystems is a significant component of the global biogeochemical silica cycle (Conley, 2002). It forms an essential buffer between mineral weathering, the ultimate source of dissolved silica (DSi), and the eventual export to the aquatic environment. Biogenic of magnitude more soluble than mineral silicates (Farmer et al., 2005) and the recycling of ASi exhibits an important control on terrestrial DSi export (e.g. Derry et al., 2005;Fulweiler et al., 2005;Sommer et al., 2006). DSi mobilization in the terrestrial environment is the primary silica source for aquatic and eventually coastal and oceanic environments (Van Cappellen, 2003), where it is an essential nutrient for diatoms  larophyceae). The availability of DSi relative to other nutrients can influence the occurrence of harmful algal blooms (e.g. Smayda, 1997). In addition, the uptake of DSi and burial of diatoms on the ocean floor is an important sink for atmospheric carbon on biological time scales (Dugdale et al., 1995). Moreover, terrestrial mineral silicate weathering is an essential controlling mechanism for atmospheric CO 2 concentrations 10 on geological timescales (Berner et al., 1983). Understanding the land-based cycling of silica is essential in understanding carbon biogeochemistry and eutrophication.
Yet, the biotic control of land-based silica fluxes is far from understood. Quantifying the ecosystem biotic component is one of the main challenges in understanding the biogeochemical cycling of silica. Research in terrestrial ecosystems has largely 15 focused on forested ecosystems, where phytolith dissolution has been identified as the primary source for exported DSi (Bartoli, 1983;Farmer et al., 2005). More recently, a literature overview suggested that wetlands have the potential to exhibit a profound influence on terrestrial silica transport and storage (Struyf and Conley, 2008). Wetlands can modify nutrient fluxes along the aquatic continuum through transformation 20 and storage. Silica rich biomass, together with the frequent occurrence of diatoms and sponges in wetlands, provides the potential for the accumulation of ASi rich soils (Clarke, 2003;Struyf et al., 2005). Struyf and Conley (2008) hypothesized that a tradeoff between the storage of ASi in wetland soils and the export of DSi from wetlands depends on hydrological and vegetation characteristics. However, little research has wetlands in reducing excessive N and P inputs to aquatic ecosystems (e.g. Hattermann et al., 2006;Olde-Venterink et al., 2006). Silica cycling related research in wetlands has focused on tidal marshes (Struyf and Conley, 2008), which are characterized by low water residence times and frequent flooding leading to the rapid recycling of buried ASi (Struyf et al., 2006), and with increasing sedimentation resulting in more efficient burial 5 of ASi (Struyf et al., 2007a). The marsh grass Phragmites australis is a key participant in the uptake of DSi, its transformation to ASi and recycling in tidal marshes (Struyf et al., 2007b). Research in tidal wetlands clearly shows the importance of hydrological connectivity and vegetation in wetland silica biogeochemistry.
We have examined the storage of ASi in a central-European riparian wetland in 10 Poland, characterized by a dominance of sedge and grass vegetation. A number of wetland plants are important accumulators of ASi (e.g. Lanning and Eleuterius, 1985;Struyf et al., 2005), but Cyperaceae (sedges) and Poaceae (grasses) are the most efficient accumulators. Wetland grasses and sedges on average contain 0.5-4% of Si in their tissues, while exceptionally high values up to 20% have been recorded (Struyf and 15 Conley, 2008). We sampled for ASi along gradients in distance to the river (as a proxy for hydrology and river connectivity), in the dominance of grasses and sedges and at both managed (annual mowing) and unmanaged sites. Porewater DSi concentrations were sampled as an indicator of recycling of ASi. Our sampling was aimed to test the hypothesis by Struyf and Conley (2008) that silica storage and recycling depends on 20 river connectivity and vegetation. Our data show that river distance and vegetation can explain most of the observed variation in ASi storage, with human management (mowing) interfering with and reducing these control mechanisms.
898 2 Materials and methods

Sampling site
The Biebrza River valley ( Fig. 1 2 cores (Ø 2 cm) of 30 cm were taken at every site in June 2007 and in total 68 cores were taken. Every core was subdivided into three sections of 10 cm (0-10 cm, 10-20 cm and 20-30 cm) and every depth section was analysed for ASi concentration. ASi was determined by alkaline extraction of 30 mg of freeze-dried sediment in 1% Na 2 CO 3 5 solution over a 5 h period with sub-samples taken at 3, 4 and 5 h and neutralized as adapted by Conley and Schelske (2002). The extract was analysed for DSi and plotted against time and the y-intercept was considered to be the ASi content to correct for simultaneous dissolution of DSi from minerals.

Vegetation
Vegetation composition at all sampling sites was determined at four randomly chosen plots using a cover pin frame (75 points/frame) in June 2006 (Kent and Coker, 1995). The combined cover of grasses (Poaceae) and sedges (Cyperaceae) was expressed as a percentage of total individuals counted, belonging to either of both plant families.

20
Dominance was preferred over number of individuals, as it was assumed to better reflect the overall vegetation composition on a longer time scale, rather than current absolute number of individuals counted. A summary of the dominant plant genera (all species with 10% or more coverage) at every site is provided in

ASi concentrations
The highest ASi concentrations were observed at sites closest to the Biebrza River and, in general, in the top 10 cm of the soil profile (Fig. 3). However, at sites near the river the highest ASi concentrations were observed in deeper layers. The areal 10 distribution of ASi in the upper 10 cm of sediment extrapolated over the study area also shows higher ASi concentrations closer to the river (Fig. 4b). General depth related trends in ASi concentration were investigated using paired two-tailed Wilcoxon signed rank tests (corrected for multiple testing), comparing depth sections within the 34 cores taken at sites with unmanaged vegetation and managed vegetation. The relationship between vegetation composition and proximity to the Biebrza River on ASi concentration was analysed for the upper 10 cm of soil. A significant logarithmic decrease of ASi concentrations in the top 10 cm of sediment with increasing distance perpendicular to the river was found both at managed (R 2 =0.29; F 1,32 =13.36; p<0.001) 5 and unmanaged vegetation sites (R 2 =0.46; F 1,32 =27.48; p<0.00001) (Fig. 5a). The relationship showing decreasing ASi concentrations with increasing distance from the river is also clear in Fig. 4b, albeit that the influence of vegetation partly masks the effect. However, there was no clear effect of combined grass and sedge dominance (or grass and sedge dominance separately, data not shown) on observed ASi concen-10 trations in the upper 10 cm (Fig. 5b). Yet, accounting for the logarithmic river distance effect on ASi concentrations and "correcting" observed ASi concentration with the "expected" ASi concentration from the distance effect, an additional effect of vegetation was found (residual ASi=observed ASi-ASi expected from distance). A positive linear relationship between combined Poaceae and Cyperaceae dominance and residual 15 distance corrected ASi is apparent in the top soil (Fig. 5c) in the unmanaged vegetation (R 2 =0.30; F 1,32 =14.00; p<0.002). A similar relationship was found in the managed vegetation ( Fig. 5c), but the ASi concentration increase with dominance was less steep, and the relationship was only marginally significant (R 2 =0.13; F 1,32 =4.78; p<0.04). It is important to note that there was no significant linear relationship between the dis-20 tance perpendicular to river and the combined grass and sedge dominance both at managed and unmanaged vegetation sites (Fig. 5d, Managed vegetation: R 2 =0.04; p>0.46, Unmanaged vegetation R 2 =0.08; p>0.28).
Combining the grass and sedge dominance and the distance to the Biebrza River in one factor (by dividing the Poaceae and Cyperaceae dominance by perpendicu-25 lar distance to the river), a significant linear relationship between this new factor and observed ASi concentration was found, both in the managed and unmanaged vegetation. Combining vegetation and distance to the river we account for (R 2 ) 78% 902 (F 1,32 =115.06; p<10 −11 , R 2 for only distance was 46%) of the observed ASi variation at the unmanaged vegetation sites, and 51% of the variation at the managed vegetation sites (F 1,32 =33.41; p<0.00001, R 2 for only distance was 29%) (Fig. 6). For both sites containing managed and unmanaged vegetation, the combined factor improves ASi variation accounted for by 60 and 70% compared to distance to the river alone. Both 5 the distance from the river and the vegetation independently account for part of the ASi variation observed over the wetland. The combined effect of vegetation and distance is stronger at unmown sites. The combined influence of vegetation and distance from the river is also apparent in Fig. 4a and b. While the gradient associated with distance from the river is situated along a North-South axis, the grass and sedge dominance 10 primarily increases from West to East. The result is highest ASi concentrations in the SE part and lowest concentrations in the NW part of the study area.

Porewater concentrations
For both managed and unmanged vegetations, highest porewater concentrations were generally observed in the sites with highest ASi (Fig. 7). A significant positive lin-15 ear relationship was actually observed between ASi concentrations (average for both observations in one site and one management type) in the top 10 cm and pooled porewater DSi concentrations at the same sites for the combined managed and unmanaged datasets (R 2 =0.45; F 1,32 =12.04; p<0.001). The association between soil ASi and porewater DSi is evident in Fig. 4b and c, where porewater DSi shows the same 20 gradient as ASi concentrations.

Discussion
We show here that the storage of ASi in riparian wetlands is significantly related to riverine connectivity and vegetation, as previously hypothesized by Struyf and Conley (2008). In this review, the authors showed that wetlands can profoundly impact the biogeochemical silica cycle through the activities of silica accumulating organisms such as diatoms, sponges, grasses and sedges, although few studies have documented this role of wetlands in silica cycling. Our data support this new understanding. In addition, we show that ASi storage in wetlands is highly variable on small spatial scales, and that recycling to DSi is directly affected by this variation. Our data suggest a link 5 between vegetation (dominance of grasses and sedges) and ASi storage. Evidence linking vegetation to storage of ASi in soils has been available from other ecosystems. In forest soils, higher ASi concentrations in soil profiles occur in forests dominated by beech (relatively rich in ASi) as compared to pine, a species poor in Si ( Bartoli and Souchier, 1978;Bartoli, 1983;Saccone et al., 2008). The largest accumulation of ASi 10 in soils has been observed at the island of Réunion from the historical accumulation of phytoliths contained in bamboo, the most efficient Si accumulator among all higher plants (Meunier et al., 1999). Blecker et al. (2006) hypothesized that storage of ASi in grasslands also strongly depends on precipitation, with high precipitation and frequent soil water renewal stimulating export of DSi from the system, after dissolution of soil 15 ASi.

The river distance proxy
In our study, river distance has been used as a proxy for hydrological characteristics and river connectivity. The proxy accounted for a significant fraction of the variation in ASi concentrations, independently from grass and sedge dominance. We hypothe-20 size that the observed variation of ASi with distance from the river can be explained by a mechanism observed previously in tidal wetlands. During winter and spring, the Biebrza River floods the Rogozynek site. During these flooding events, the river imports suspended material into the floodplain, potentially co-importing ASi as both phytoliths and diatom frustules in suspended particulate matter. Within the wetland, stream 25 velocities quickly drop, causing all but the lightest particles to be deposited near the river. In tidal marshes, it has been shown that most suspended matter is deposited close to the river or creek banks (Temmerman et al., 2003), and a similar process 904 was observed in non-tidal floodplains (Asselman and Middelkoop, 1995). Dense vegetation, as observed along the Biebrza, further stimulates rapid settling of particulate suspended matter. The highest import of allochtonous ASi is therefore expected near the river. Another possible mechanism for the enhanced ASi concentrations near the river, would be a tendency for more Si rich grasses and/or sedges to occur. The site 5 richest in ASi (site 9) was characterized by a relatively important dominance of Phragmites australis, well known for its large content of ASi compared to most other grasses (Hodson et al., 2002;Struyf et al., 2005). Unfortunately, we have no data of ASi in aboveground biomass for the Rogozynek site.
Storage of ASi could also be directly related to groundwater hydrology. Frequent 10 oxidation-reduction cycles has been proposed to stimulate the dissolution of ASi (Beckwith and Reeve, 1963;Sommer et al., 2006) with the sorption of DSi into sesquioxides under oxygenated conditions and adsorbed DSi subsequently released after a return to reduced conditions. We have no full hydrological characterization of the site, and cannot confirm if oxidation-reduction is more intense at sites farther from the river.
However, at least one piezometer near to the river showed persistent reduced conditions. Permanent water-logging could also slow dissolution of ASi for long time-periods through the build-up of DSi concentrations in the soil water. At sites with more frequent flooding and flushing of porewater with a greater flow of water, enhanced dissolution would be expected. Finally, more permanent wet conditions could enhance the growth 20 of diatoms and sponges enhancing the accumulation of ASi. It is clear that the link between ASi storage and recycling in wetlands and hydrology needs further study to quantify its impact on the small-scale variation observed.

Vegetation and management
Our results demonstrate that the pre-dominance of ASi rich vegetation can significantly 25 stimulate ASi storage in wetland soils, as previously shown on larger spatial scales in different ecosystems (e.g. Bartoli, 1983;Jensen et al., 2003). This would imply that Poaceae and Cyperaceae dominance, as frequently observed in wetlands, increases 905 DSi availability in soils available to plants by building a recyclable pool of soil ASi. Increased DSi availability through dissolution of ASi could in turn enhance grasses' and sedges' competitiveness. DSi uptake by plants is well known to enhance their resistance to multiple stressors, and thus potentially their competitiveness (e.g. Epstein, 1999;Massey et al., 2007). We hypothesize that ASi build-up as a result of biomass 5 burial, and resulting high porewater DSi concentrations, might reinforce DSi availability and grasses' dominant position, resulting in a positive feedback loop. A continuous import of DSi from the river or from groundwater is a necessary condition for this hypothesis of surficial ASi build-up and should be fulfilled in a majority of riparian wetland sites, buffering groundwater and surface water fluxes between terrestrial and aquatic 10 environments. Riverine connectivity and vegetation accounted for more of the observed variation in ASi concentrations at sites not managed by mowing. This is coherent with increased trapping of allochtonous material in dense tussock vegetation and increased grass biomass burial in the absence of mowing. The ASi concentrations in our study site 15 were low (<1% by weight), yet control of ASi on DSi concentrations was apparent. ASi is well-known for the control it exhibits on soil DSi concentrations (Farmer et al., 2005;Struyf et al., 2005), as it is more soluble than mineral silicates (Van Cappellen, 2003). The depth gradient of preservation we observed also indicates dissolution of ASi with depth. The results imply that wetland management and its impact on ASi 20 storage in wetlands, can consequently alter DSi available in the soil water. Export of water from the wetland to the aquatic environment mainly occurs as advective leakage to the river of soil water, especially during rain events, or through vertical leakage to the groundwater. As such, ASi dissolution probably controls export of DSi from the riparian habitat into the river system. In estuaries, wetlands are buffers in the silica 25 cycle, exporting Si after recycling of buried ASi in times of Si-depletion (Struyf et al., 2006;Jacobs et al., 2008). Porewater DSi is the main source for the DSi export (Struyf et al., 2006). We suggest that a similar mechanism occurs in non-tidal wetlands.

906
can significantly influence ASi dynamics within wetlands over small spatial scales. The relative importance of local hydrology vs. allochtonous import of material could not be distinguished in this study, as distance from the river could be directly linked to both (although the presented data suggest that groundwater hydrology is rather uniform over the whole site). Future work should further constrain and identify these factors. Our understanding of the importance of wetlands to riverine Si biogeochemistry and the transport of Si along the land-ocean continuum is growing. To improve our understanding of both the sink and source functions, exchange studies in a variety of wetlands are a necessary prerequisite, as well as enhanced understanding of wetland Si dynamics.
Our results indicate that vegetation and hydrology should be included as forcing fac-15 tors in these studies. The importance of riverine Si transport in coastal productivity, where Si, N and P exhibit important controls on phytoplankton composition and productivity, certainly warrants a scientific effort that would enable our understanding of Si biogeochemistry to increase to the level of understanding achieved for N and P cycling. Geochem., 54, 357-381, 2003