Journal list menu

Volume 65, Issue 2 p. 401-412
Article
Free Access

How progressive vegetation die-off in a tidal marsh would affect flow and sedimentation patterns: A field demonstration

Lennert Schepers

Corresponding Author

Lennert Schepers

Ecosystem Management Research Group, University of Antwerp, Wilrijk, Belgium

Correspondence: [email protected]Search for more papers by this author
Alexander Van Braeckel

Alexander Van Braeckel

Team Estuaries, Research Institute for Nature and Forest (INBO), Brussels, Belgium

Search for more papers by this author
Tjeerd J. Bouma

Tjeerd J. Bouma

Department of Estuarine and Delta Systems, Royal Netherlands Institute for Sea Research (NIOZ), Yerseke, The Netherlands

Search for more papers by this author
Stijn Temmerman

Stijn Temmerman

Ecosystem Management Research Group, University of Antwerp, Wilrijk, Belgium

Search for more papers by this author
First published: 26 August 2019
Citations: 6
Associate editor: Josef Ackerman

Abstract

Coastal marshes provide valuable ecosystem functions, but some are facing increasing risks of vegetation loss due to sea level rise and other stressors. A key question is how tidal flow and sedimentation patterns are affected by the spatiotemporal patterns of vegetation loss, as sediment accretion with sea level rise largely affects the potential for marsh recovery. Here, we performed a field study in a macrotidal reed marsh and simulated typical spatiotemporal patterns of vegetation loss by consecutive mowing. For each mowing pattern, the spatial patterns of flow velocities and sedimentation rates were recorded. Our results indicate that initial vegetation loss in inner marsh portions, with an intact vegetation belt alongside channel edges, has limited effect on tidal flows over the marsh. However, subsequent creation of unvegetated corridors connecting the bare inner marsh and the channels increases flow velocities in these corridors but not in remaining vegetation patches. Finally, when all vegetation is removed, sheet flow occurs over the whole marsh instead of concentrated channel flow. Effects on spatial sedimentation patterns are complex and not significant on all measuring locations. Nevertheless, our study indicates that complete vegetation removal results in redistributed sedimentation patterns, with a general tendency of locally reduced sedimentation rates close (< 15 m) to channels and increased sediment supply to inner marshes 15–50 m from channels. Our results highlight that feedbacks between spatial patterns of vegetation loss, tidal sediment transport, and deposition are key to understanding and mitigating risks of marsh loss in face of sea level rise.

Sea level rise is challenging the persistence of coastal marshes and their valuable ecosystem services, including carbon sequestration (McLeod et al. 2011; Duarte et al. 2013), attenuation of storm surges (Smolders et al. 2015; Stark et al. 2015), and storm waves (Möller 2006; Gedan et al. 2011; Möller et al. 2014). Marsh vegetation plays an important role in the provision of these services and in the survival of coastal marshes with sea level rise, as the vegetation reduces waves and flow velocities and promotes sediment deposition on the marsh surface (Bouma et al. 2005, 2010; Baustian et al. 2012), enabling many marsh areas to keep up with increasing sea level rise (Kirwan et al. 2016). However, if sediment deposition is not sufficient, sea level rise may provoke increasing marsh flooding until it exceeds the marsh plant's tolerance for inundation, ultimately leading to a loss of marsh vegetation (Kirwan et al. 2010, 2013). A key question is then how marsh loss affects flow and sedimentation patterns, as this may induce detrimental consequences for the ecosystem functions of marshes.

Complete removal of vegetation increases flow velocities on the marsh platform, which has been shown in a field experiment (Temmerman et al. 2012) and in a number of modeling studies (Temmerman et al. 2005; Ashall et al. 2016; Wu et al. 2017). The increase in flow velocity has been attributed to reduced friction and would imply lower sedimentation rates or even erosion after marsh die-off (Silliman et al. 2012; Sheehan and Ellison 2015; Coleman and Kirwan 2019). In the unvegetated channels that typically cut through marsh platforms and that supply water and sediments to and from the surrounding marsh platform, a contrasting effect has been demonstrated in a field experiment: complete removal of surrounding marsh vegetation resulted in lower channel flow velocities (Temmerman et al. 2012). Comparison with modeling showed that canopy drag of the marsh vegetation typically concentrates flood and ebb flows toward the channels where friction is lowest; vice versa, complete vegetation removal leads to lower flow velocities (Temmerman et al. 2005; Ashall et al. 2016) and more sedimentation in the channels, which might even lead to partial channel infilling (Temmerman et al. 2005). In agreement with this vegetation effect on channel flow and sediment transport, aerial photo analyses have demonstrated that channel networks are denser (more closely spaced) in vegetated intertidal marshes as compared to nonvegetated intertidal flats (Temmerman et al. 2007; Vandenbruwaene et al. 2013; Kearney and Fagherazzi 2016).

However, to the best of our knowledge, the only two experimental studies of large-scale vegetation removal (Temmerman et al. 2012; Sheehan and Ellison 2015) considered complete, instantaneous vegetation removal, whereas marsh vegetation die-off is a gradual process with distinct spatiotemporal patterns of vegetation loss (Mariotti 2016; Schepers et al. 2017). Initial die-off typically starts at the inner portions of marshes at a distance from channels, where surface elevation is lower and soil drainage is less developed as compared to the higher elevated natural levees directly adjacent to channels (Redfield 1972; Kearney et al. 1988; Schepers et al. 2017). These areas with initial marsh die-off expand and subsequently become connected to the tidal channels network. Connections with the tidal channels can lead to sediment infilling and vegetation recovery (Perillo et al. 1996; Wilson et al. 2010, 2014), whereas other research suggests that connections can lead to erosion of the bare inner portions of marshes (Day et al. 2011; Schepers et al. 2017). Hence, the spatiotemporal patterns of vegetation die-off are expected to have a crucial impact on the tidal flow and the spatial sedimentation pattern. Yet, no field studies have directly quantified the effects of the different stages of vegetation loss on tidal flow and sedimentation patterns.

In this study, we performed a large-scale field study in which the different stages of tidal marsh vegetation die-off are simulated by artificial mowing of the vegetation in a sequence of spatial patterns with reducing vegetation cover. Sediment deposition and tidal flows were measured and repeated for the consecutive mowing patterns. We hypothesize that removal of vegetation in inner parts of the marsh at a distance from the channels would have limited impact due to the intact vegetation buffer along the channels. In contrast, further vegetation removal connected to channels as well as complete vegetation removal would increase flow velocities significantly and limit sediment deposition on the marsh platform.

Methods

Study area

The field study was conducted in a 16 × 103 m2 area within a freshwater tidal marsh (Kijkverdriet) in the Scheldt Estuary, Belgium (Fig. 1, 51.1213°N, 4.2641°E). The semidiurnal tidal range is 6 m during spring tides and 4.5 m during neap tides. The geomorphology of the minerogenic marsh consists of a generally flat marsh platform with an elevation of 0.05 ± 0.17 m (mean ± SD) relative to mean high water level (Temmerman et al. 2012). The marsh platform is dissected by a single unvegetated, 2.5 m wide, 1.4 m deep tidal creek (Temmerman et al. 2012) (Fig. 1B,C). The marsh consists of a uniform Phragmites australis (common reed) vegetation up to 4 m high, which stays emergent even during the highest spring tides. Local suspended sediment concentrations in the adjacent estuarine channel range between < 5 and 300 mg L−1 (Temmerman et al. 2003a). At the location of Kijkverdriet, the Scheldt river is about 350 m wide (Fig. 1B) and peak ebb and flood velocities in the river channel are higher than 1 m s−1 (Plancke et al. 2009). As such, because of the small wind fetch length, shallow water depths on the marsh (< 0.7 m; Fig. 2), and large tidal range (4.5–6 m), we can consider that tidal currents are the dominant hydrodynamic forcing while wind-induced waves and currents are negligible at this location. Salinities range between 0.4 and 2.1 (data: http://www.scheldemonitor.be/, 25 November 2017).

Details are in the caption following the image
(A) Overview of the Scheldt Estuary with the location of the field study indicated as (B). (B) Digital terrain model of the Kijkverdriet marsh (Informatie Vlaanderen 2014). (C) Overview of the study area with locations of flow velocity measurements (red dots) on the marsh platform (A–C) and in the tidal creek (D–E), water level measurements (black dot), sediment traps (white dots labeled 1.1–4.4), and mowing patterns (different shades of green): Pattern 1 = “fully vegetated” = with all vegetation present (i.e., in all zones with different shades of green); pattern 2 = “inner mown” = with vegetation removal on the inner portion of the marsh (i.e., zones indicated with lightest shade of green); pattern 3 = “connections mown” = with vegetation removal on the inner portions (i.e., lightest shade of green) and in connections with the creek and river (i.e., darkest shade of green); pattern 4 = “all mown” = complete vegetation removal (i.e., in all zones with different shades of green). P indicates the approximate location from which Supporting Information Fig. S1 was taken (see Supporting Information).
Details are in the caption following the image
Selected tides for the flow velocity measurements, with very similar shapes of the hydrographs and with high water levels between 0.40 and 0.75 m above mean marsh elevation.

Flow and sedimentation measurements

The vertical tide was measured using a pressure sensor (diver, Eijkelkamp) recording water levels at the mudflat just in front of the main tidal creek every 30 s. The water levels were corrected for barometric pressure variations with a simultaneously deployed pressure sensor outside the marsh (baro-diver, Eijkelkamp). We recorded the elevation of the sensor with a high-precision GPS (Trimble R4 RTK-GPS, vertical error < 1.5 cm) to refer the water level measurements to the mean marsh platform elevation as measured by Temmerman et al. (2012).

We measured tidal flow and sedimentation patterns at different locations on the marsh surface. Tidal flows were measured in the creek (locations D and E in Fig. 1C) and on the marsh platform (locations A, B, and C in Fig. 1C). In the creek, two high-resolution acoustic Doppler velocity profilers (HR-ADCPs, Nortek AS) were attached to the bottom of the creek and deployed upward-looking. On the marsh platform, three acoustic Doppler velocity meters (ADV, Nortek AS) measured three dimensional flow velocities at 12 cm above the marsh platform. The marsh vegetation was left as intact as possible, but we removed stems within 30 cm surrounding the ADV measurement volume that could possibly interfere with the ADV sonar beams. One ADV was positioned near the riverside edge of the marsh, one near the edge of the main creek, and one at the inner marsh platform (Fig. 1C).

During each tide, a sample of the flooding water was collected using a siphon sampler (1-liter bottle with siphon tubes that is filled when one of the tubes is submerged). The sample was taken at the creek inlet, at the average marsh elevation. The suspended sediment concentration from this floodwater was determined by filtering and weighing preweighed filter papers (240 mm Whatman Hardened Low Ash filter papers).

Sediment deposition during single tides was measured on 16 spatially distributed, circular sediment traps (Fig. 1C). The design and validation of the sediment traps is presented in detail in Schoelynck et al. (2015). In brief, they consist of PVC plates that were fixed to the sediment surface and that held circular filter papers (diameter 0.2 m), onto which suspended sediment deposition takes place. This method is a measure of sedimentation and resuspension of the deposited sediment; it cannot measure net erosion. Nevertheless, this method is currently the best and most accurate method for determining sedimentation over single tides (Nolte et al. 2013; Schoelynck et al. 2015). We used the sedimentation data for relative comparison between different mowing patterns. Filter papers were dried and preweighed at 105°C and labeled before applying them in the field. After each high tide, the traps were collected, dried at 105°C to constant weight to determine the weight of the sediment deposited on the filter papers.

Mowing patterns

Marsh vegetation die-off usually starts at the interior parts of marshes at a distance from channels and creeks (Stevenson et al. 1985; DeLaune et al. 1994; Schepers et al. 2017). These vegetation-loss areas may subsequently become larger and may connect to the tidal channel network (Redfield 1972; Kearney et al. 1988; Schepers et al. 2017). We simulated the effect of subsequent stages of marsh loss by consecutive reed mowing, after which only short stems of maximum 0.1 m high remained on the marsh platform. We first measured flow velocities and sedimentation rates on the fully vegetated marsh (no mowing applied) during October–November 2013. Before the measurements in December 2013, the inner marsh platform was mown (Fig. 1C, light green areas), leaving a vegetated buffer of 15 m along both sides of the creek and along the riverside edge of the marsh. As a second stage, vegetation-free connections were mown from the river and the creek to the interior marsh (Fig. 1C, dark green areas). Measurements with this pattern were carried out in February and March 2014 (Supporting Information Table S1). Since we could not finish our measurements in one season, the fully mown stage was measured at the end of the next winter season in February 2015 (mowing occurs in winter for reed and target plant species conservation such as spring-flowering Caltha palustris subsp. araneosa, Leucojum aestivum). All mowing patterns were applied symmetrically at both sides of the tidal creek (Fig. 1C).

Data processing and analysis

In order to compare flow measurements between similar tides, we only considered tides with high water levels between 0.40 and 0.75 m above the average marsh platform elevation and with very similar shapes of the hydrograph (Fig. 2, Supporting Information Table S1).

Creek HR-ADCP data were filtered to retain only data with amplitudes higher than 50 counts (1 count is approximately 0.4 dB) and correlations higher than 50%. We corrected the HR-ADCP data for velocity folding. This happens when the recorded velocity is higher than the predefined velocity range resulting in a phase shift and incorrect velocity estimates (Franca and Lemmin 2006). This is a known issue and can easily be corrected by adding or subtracting two times the maximum velocity (e.g., see Supporting Information Fig. S2). After correction, beam velocities of the extended velocity cell (0.6 m above the HR-ADCPs) were transformed to two-dimensional (2D) horizontal flow vectors at one measurement cell 50 cm below the marsh surface surrounding the creek. Marsh ADV measurements with amplitudes lower than 90 counts and correlations lower than 70% were discarded. The ADV beam velocities were also transformed to 2D horizontal flow vectors. All HR-ADCP and ADV 2D vectors were averaged per minute, and from these the flow velocity and flow direction were calculated. All measurements were grouped by location, tidal stage (flood: 2 h before high water, ebb: 2 h after high water), and mowing pattern (Table 1). The moment of high water was determined by the water level sensor at the river. Velocity differences between the mowing patterns at each location were tested by the nonparametric pairwise Wilcoxon rank sum test with Bonferroni correction (α = 0.05).

Table 1. The number of values (1-min flow averages) used in the flow analysis. The numbers are given for each tidal stage (flood/ebb), at each location (rows) and mowing pattern (columns).
Flood/ebb 1: Fully vegetated 2: Inner mown 3: Connections mown 4: All mown
Location A: ADV 229/421 131/248 147/303 174/274
Location B: ADV 219/276 143/177 280/381 163/208
Location C: ADV 219/271 126/152 243/312 168/185
Location D: ADCP 770/865 535/702 1173/1656 585/884
Location E: ADCP 686/761 540/725 1173/1687 595/886

Due to rain-splashing, we could not use the sediment traps of several tides (Supporting Information Table S1). Since sediment deposition is determined by incoming suspended sediment concentration, which varies significantly even on short temporal timescales (Fettweis et al. 1998; Temmerman et al. 2003b; Butzeck et al. 2014), we quantified the relative spatial sedimentation patterns for each tide, calculated as the ratio between the local sedimentation rate measured at a point and the spatially averaged sedimentation rate of the 16 sediment traps measured during the same tide (i.e., a relative sedimentation rate of 0.5 indicates that the sedimentation rate at that location was half of the spatially averaged sedimentation rate for that tide). We compared the spatial sedimentation patterns between different tides. Significant differences between the mowing patterns at each location were tested by the nonparametric pairwise Wilcoxon rank sum test with Bonferroni correction (α = 0.05). All analyses were performed in R (R Core Team 2017).

Results

Effect of mowing on flow velocities

For the vegetated treatment (i.e., fully vegetated marsh platform, Fig. 1C), the flow velocities measured on the vegetated marsh platform were low (mostly < 0.05 m s−1) but were higher within the nonvegetated tidal creek (up to ca. 0.5 m s−1 during peak flood and almost reaching 0.8 m s−1 during peak ebb) (Fig. 3). Compared to the vegetated treatment, the conditions in the mowninner pattern did not significantly affect the average ebb or flood velocities on the interior marsh platform that was mown (Fig. 3A), and did not significantly affect the flood velocities on the creek edges that remained vegetated (Fig. 3B,C). It did have a significant effect on flood and ebb velocities measured in the creek (Fig. 3D,E). For mownconnected, parts of the vegetation buffer bordering the creek and the river were additionally removed (Fig. 1C). This impacted the overall flow patterns as the mown connections became preferential flow routes (Fig. 4). In particular at the marsh platform location near the river (location C in Figs. 2-4), which was still vegetated in mowninner but was in a mown connection in mownconnected (Fig. 1C), both the flood and ebb velocities significantly increased from mostly < 0.05 m s−1 for the vegetated and mowninner pattern (both with vegetation at location C) to almost 0.2 m s−1 during peak flood and ebb for mownconnected (mown connection at location C) (Fig. 3C). On the interior marsh platform (location A), the mowing of the connections had only a limited but significant effect on slightly higher flood velocities, and no significant effect on ebb velocities (Fig. 3A). Within the remaining vegetation bordering the creek (location B), no significant changes in flow velocity or direction were detected between mowninner and mownconnected (Fig. 3B, Supporting Information Fig. S3). Changes in channel flow velocities between these two patterns were mostly not significant (Fig. 3D,E).

Details are in the caption following the image
Boxplots of 1-min averaged flow velocities for the different mowing patterns (x axis) and the different measurement locations (see Fig. 1C): (A) Inner marsh platform, (B) on the platform near the main creek, (C) on the platform near the river, (D) in the creek upstream, and (E) in the creek closest to the river. Significantly different velocities between different mowing patterns at each location are indicated by different letters. Differences between the velocities of different mowing patterns were tested separately at each location with the pairwise Wilcoxon rank sum test with Bonferroni correction (α = 0.05). Light gray background means that the measuring point is within the vegetation.
Details are in the caption following the image
Flow velocities and directions on the marsh change with different mowing patterns. Vegetated zones are indicated in dark green, mown zones in light green. Mowing vegetation-free connections between the mown inner marsh and the tidal creek increases the flow in these corridors. When the marsh is fully mown, the flood direction and the ebb direction near the creek flow in the same direction, from the marsh surface toward the tidal creek. See Supporting Information Fig. S3 for the creek measurements.

Complete mowing (mownall) did not further increase the platform flood velocities significantly, and it resulted in a small but significant increase of ebb velocities at two of the three platform locations (Fig. 3A,B). Moreover, complete mowing had a visible effect on the flow direction at the platform locations (Fig. 4). Near the marsh edge close to the river (location C), flood and ebb flows were no longer perpendicular to the marsh edge, but now the dominant water flow direction was parallel to the riverside marsh edge (Fig. 4; although the direction also repeatedly changed, resulting in a broad range of flow directions). At location B near the creek, the flood flow was no longer oriented perpendicular from the creek edge toward the marsh platform, but instead the flood flow was completely opposite from the marsh platform toward the creek. This flow direction and the dominant flow direction on all three platform locations were similar and more or less parallel with the riverside marsh edge, hence suggesting that there was large-scale sheet flow from the riverside marsh edge over the nonvegetated marsh platform (Fig. 4).

In the tidal creek, the different mowing patterns appear to have relatively little and not consistent effects on the flow velocity (Fig. 3D,E; Supporting Information Fig. S3D,E).

Effect of mowing on sedimentation patterns

Suspended sediment concentrations varied considerably between seasons and individual tides (ranging between 65 and 325 mg L−1) during the measuring period. Hence, we calculated the relative spatial variations in sedimentation rates, i.e., the local sedimentation rate divided by the average of all sediment traps per individual tide.

The different mowing patterns showed spatially complex effects on the sedimentation patterns (Fig. 5, Supporting Information Fig. S4), but some trends are not consistent among the measurement locations. The most evident and consistent effects are the differences between the fully vegetated marsh and the fully mown marsh (Fig. 5; data on all intermediate vegetation patterns are analyzed in Supporting Information Fig. S4). In the fully vegetated marsh case (purple/dark bars in Fig. 5), all locations bordering the river and tidal creek experienced relative sediment rates exceeding the spatially averaged rate. The inner marsh locations experienced lower than average sedimentation rates. When the vegetation was completely removed, a contrasting spatial sedimentation pattern became apparent (yellow/light bars in Fig. 5): relative sedimentation rates were lower than the spatial average on all locations bordering the creek, while on most inner marsh locations, relative sedimentation rates were larger in the fully mown case as compared to the fully vegetated case. The relative sedimentation rates were significantly different at five of the locations (white asterisks on Fig. 5). At three out of four locations bordering the river, differences between the fully vegetated and mown situations were not significantly different (Fig. 5).

Details are in the caption following the image
Relative sedimentation rates, calculated for each measuring location as the location-specific sedimentation rate divided by the spatially averaged sedimentation rate (i.e., averaged over all measuring locations) during a given individual tidal inundation event. At each measuring location (white points on the map), the mean relative sediment rates (indicated by bars) and standard error (indicated by white vertical lines) are plotted for all individual tides with full vegetation cover (purple/dark bars, n = 9) and all tides with all vegetation mown (yellow/light bars, n = 5) for reference, and the horizontal (red) lines indicate the spatially averaged sedimentation rate averaged over all tides. White asterisks above the bar graphs indicate significant differences (Wilcoxon rank sum test) in relative sedimentation rates between the fully vegetated and fully mown situations. More sediment is deposited close to the creek compared to the inner marsh with full vegetation, while the opposite is true for the mown marsh.

Discussion

Effect of mowing on flow velocities and directions

Our results demonstrated that (1) mowing of the inner marsh vegetation while keeping a continuous vegetation belt alongside the creek and river (mowninner) did not significantly affect flow velocities at the inner marsh platform; (2) additional mowing of corridors through the vegetation belt, connecting the river to the inner (mown) marsh (mownconnected), resulted in a significant, drastic increase of flow velocities in the mown corridor but (3) not within the remaining vegetation bordering the creek; (4) complete vegetation removal (mownall) resulted in changed flow directions, from flow perpendicular to and from the creek edge and riverside marsh edge to large-scale sheet flow over the bare marsh platform directed parallel to the large-scale river flow direction; and (5) all mowing patterns had relatively small and inconsistent effects on creek velocities.

The fact that the flow velocities are not significantly affected when vegetation is only removed in the inner marsh (mowninner) (Fig. 3A–C) can be explained as the remaining vegetation buffer alongside the river and creek edges still reduces the flow velocities as soon as the flow enters from the river and creek onto the marsh platform. This explanation is consistent with previous field studies that show the strong reduction of flow velocities within the first few meters of marsh vegetation (Christiansen et al. 2000;Leonard and Croft 2006 ; Mudd et al. 2010). Leonard and Croft (2006), for example, found that 50% of the mean velocity was dissipated within 5 m of a Spartina alterniflora canopy. Thus, even a small buffer (15 m) or small vegetated patches can have a profound effect and reduce flow velocities significantly.

The increased flow velocities in nonvegetated corridors connecting the river with the nonvegetated inner marsh are likely due to flow routing and acceleration between the remaining vegetation canopy. This effect of flow acceleration around and in between vegetated patches is well known for the case of pioneer vegetation patches on low-elevation tidal flats, as demonstrated by field (Temmerman et al. 2005; Vandenbruwaene et al. 2015), flume (Vandenbruwaene et al. 2011; Bouma et al. 2013; Meire et al. 2014), and modeling studies (Temmerman et al. 2005; Meire et al. 2014). In our present study, a similar flow accelerating effect is observed in a high-elevation marsh in the nonvegetated corridor in between the remaining marsh vegetation.

Complete mowing changed the flow patterns on the marsh platform to landward sheet flow (Fig. 4, Supporting Information Fig. S3). Especially noteworthy is that near the creek, the flood flow no longer originated from the nearby creek, but instead the sheet flow from the riverside marsh edge redirected the flow toward the creek in the same direction as the ebb flow (Fig. 4). A previous study at the same location with complete vegetation removal did not show this sheet flow toward the creek (Temmerman et al. 2012), although complete mowing changed the flow velocity to a more parallel direction with the creek. The discrepancy between this previous study and our present study is probably a result of different calculation methods as (Temmerman et al. 2012) only extracted a short 15-min interval at beginning of flood and end of ebb to calculate flow directions, while here we show the flow directions during the whole inundation period. Our measurements provide for the first time empirical evidence that a nonvegetated marsh platform is flooded by sheet flow from the marsh edge instead of flooding via tidal creeks in case of a vegetated platform (Temmerman et al. 2005; Ashall et al. 2016; Wu et al. 2017).

The different mowing patterns appear to have little effect on the overall flow velocity in the tidal creek. A fully vegetated platform and a fully mown platform result in similar creek flood velocities (Fig. 3D,E, Supporting Information Fig. S3, rows D and E). Our creek velocities are comparable to the values reported in earlier studies in the same study area for the situation with the fully vegetated platform (up to 0.6 m s−1), but do not decrease with vegetation removal. This is contradictory to a previous field experiment (Temmerman et al. 2012) and modeling study (Temmerman et al. 2005) that show a reduction in creek flow velocity when marsh vegetation is removed. Ashall et al. (2016), however, found also minor influence of vegetation on creek velocities in their model study.

Effect of mowing on sedimentation patterns

The effect of different vegetation patterns on sedimentation rates is spatially complex (Supporting Information Fig. S4). Nevertheless, we could in general identify consistent patterns in different sedimentation rates between the fully vegetated marsh and the fully mown marsh (Fig. 5). In the fully vegetated marsh, the locations bordering the river and tidal creek experienced sedimentation rates that were higher than spatially averaged rates, while inner marsh locations experienced lower than average sedimentation rates. When the vegetation was fully mown, a contrasting spatial sedimentation pattern was observed, with sedimentation rates being lower alongside the creek and higher in the inner marsh (Fig. 5). This may be counterintuitive, since vegetation loss is commonly expected to result in lower sediment deposition rates in tidal marshes, which is both suggested by field studies (e.g., Baustian et al. 2012) and numerical models of marsh evolution (D'Alpaos et al. 2007; Kirwan and Murray 2007; Mudd et al. 2010). However, the flow velocity measurements can explain this remarkable result. With a fully vegetated marsh, the vegetation-induced drag on the marsh platform concentrates the tidal flow toward the nonvegetated tidal creek, while keeping the flow velocities low on the vegetated marsh platform (Fig. 4). This results in slightly faster flood wave propagation through the creek as compared to propagation over the platform, therefore resulting in a water surface sloping from the creek toward the platform and hence causing flood flow perpendicular to the creek edge toward the platform (Temmerman et al. 2005). Flow velocities are reduced within the first meters of the vegetated platform and most sediment is deposited at short distance from the creek edge (Leonard and Croft 2006; Mudd et al. 2010). Without vegetation, the marsh is mainly flooded as sheet flow from the riverside marsh edge and the sediment is transported further onto the marsh platform before it is deposited. Hence, complete vegetation loss does not simply lead to reduced sedimentation rates all over the marsh platform, but in a spatial redistribution of the sediment, resulting in locally reduced sedimentation rates close to channels and creeks, and locally increased sedimentation rates in inner marsh areas at farther distance from channels and creeks (Temmerman et al. 2005).

Intermediate vegetation removal (mowninner and mownconnected) has inconsistent and complex effects on spatial sedimentation patterns (Supporting Information Fig. S4). This is probably a result of complex flow routing patterns and is difficult to explain because we do not know for every location if the water and sediment supplied to that location was routed through the mown corridors or through the remaining vegetation. In any case, comparison of the fully vegetated vs. fully mown situations indicates that flow and sediment routing through vegetated portions of the marsh platform results in higher sedimentation rates closer to the creek and lower sedimentation rates at farther distance from the creek, while vice versa, flow and sediment routing over nonvegetated marsh surfaces results in reduced sedimentation rates close the creek and increased sediment supply and deposition farther away from the creek.

Implications for marsh die-off

Our findings may have implications for areas that experience large-scale marsh loss, including microtidal marshes lacking sufficient suspended sediment import and macrotidal marshes like ours with relatively high suspended sediment concentrations up to 325 mg L−1. The presence of vegetation on the marsh platform induces higher sedimentation rates close to creeks or river channels as compared to the inner marsh, while complete vegetation loss implies lower sedimentation rates close to creeks and more sediment deposition further into the marsh platform (Fig. 5). Several studies have shown that tidal marsh vegetation die-off, resulting from sea level rise or other stressors, typically starts at inner marsh basins (Redfield 1972; Kearney et al. 1988; Schepers et al. 2017). This implies that inner marsh areas are the typical critical zones where sediment import and deposition rates need to be high enough to keep up with sea level rise. Our results suggest that complete vegetation die-off might facilitate the transport and deposition of sediments on inner marsh parts after which marsh plants could reestablish. However, complete vegetation die-off over large areas is not an instantaneous process, but marsh die-back mostly develops gradually over many years to decades, whereby marsh vegetation typically dies off first in inner marsh portions and is able to survive for many years on the higher elevated and better drained levees bordering creeks and channels (e.g., Schepers et al. 2017). Our study suggests that these vegetated buffers alongside creeks and channels hinder the transport of sediments to the bare inner marsh parts. Once these inner marsh parts get connected to the channel network, and in areas with sufficient sediment import, die-off areas may indeed recover after sediment infilling in certain places (Wilson et al. 2010, 2014; Elschot et al. 2017).

Our results may also have implications for ecological management of tidal marshes that are at risk for die-off in inner marsh portions. Winter mowing of vegetation on the levees bordering creeks and channels could be a management tool to increase sediment supply and deposition in inner marsh portions. Where mowing is not feasible, prescribed burning or livestock grazing could be applied, as is commonly applied in the United States and NW Europe (Nyman and Chabreck 1995; Esselink et al. 2000). However, an increase in flow velocity resulting from vegetation removal could also result in more export of easily eroded soils (Stevenson et al. 1985; Wilson et al. 2010; Day et al. 2011), as a result of the higher ebb velocities (Fig. 3B,C). Future research should verify these hypotheses by in situ flow, erodibility, and sediment flux measurements in the specific areas that are at risk of large-scale marsh loss.

The ultimate recovery or degradation of the marshes depends on where and how much sediment is deposited on the marsh surface. Our structured mowing patterns (Fig. 1) show the complexity of the resulting sediment transport and spatial sedimentation patterns. Deeper understanding of the interactions between the spatiotemporal development of vegetation die-off and of sediment transport and deposition is needed to correctly predict which of the remaining marshes in submerging coastal areas will be able to survive ongoing sea level rise, or to determine where to focus restoration efforts. Future field and modeling studies on hydrodynamic and sediment transport in marsh die-off areas could contribute to deeper insights in support of sustainable management of these stressed environments.

Conclusion

We performed a large-scale field study in which consecutive stages of coastal marsh vegetation die-off were simulated by different mowing patterns. We showed that initial vegetation loss in inner marshes, with vegetation still bordering the tidal channels, resulted in limited increases of the flow velocities in the inner marsh. However, mowing unvegetated corridors between the bare inner marsh and channels leads to locally higher velocities over the unvegetated surfaces. The flow velocities in remaining vegetation patches remained unaffected. With complete vegetation removal, sheet flow over the marsh surface replaced flooding through the channels, although channel velocities were not systematically affected. Effects on spatial sedimentation patterns were complex. Nevertheless, our study indicated that complete vegetation loss did not simply lead to reduced sedimentation rates everywhere on the marsh platform. In contrast, complete vegetation loss resulted in redistributed sedimentation patterns, with a general tendency of reduced sedimentation rates at short distances from channels and increased sediment supply and increased sedimentation rates in inner marshes at farther distance from channels.

Acknowledgments

We would like to thank the early voluntary managers of the Kijkverdriet marsh for their efforts, the current conservator Joris Goossens and his team from the Flemish Agency for Nature and Forests for their assistance and support in mowing the marsh in different patterns. We thank Tom de Dobbelaer for the valuable field support as part of his master's program. This project was financed by UA-BOF DOCPRO grant (to L.S. and S.T.) and the Research Foundation Flanders (FWO, Ph.D. grants L.S., 11S9614N and 11S9616N).

    Conflict of Interest

    None declared.