There is an increasing demand for the creation and restoration of tidal marshes around the world, as they provide highly valued ecosystem services. Yet restored tidal marshes are strongly vulnerable to factors such as sea level rise and declining sediment supply. How fast the restored ecosystem
develops, how resilient it is to sea level rise, and how this can be steered by restoration design are key questions that are typically challenging to assess due to the complex biogeomorphic feedback processes involved. In this paper, we apply a biogeomorphic model to a specific tidal-marsh restoration project planned by dike breaching. Our modeling approach integrates tidal hydrodynamics, sediment transport, and vegetation dynamics, accounting for relevant fine-scale flow–vegetation interactions (less than 1
Tidal marshes are among the most productive ecosystems on Earth (Barbier et al., 2011) providing invaluable services such as the protection of coastal settlements against storms (Gedan et al., 2011; Zhu et al. 2020), carbon sequestration (Rogers et al., 2019), maintenance of fisheries (Boesch and Turner, 1984), and water purification (Breaux et al., 1995). They are however among the most threatened ecosystems globally (Barbier et al., 2011). Over centuries, humans have built dikes to prevent tidal flooding and drained soils to gain land for agricultural, industrial, and urban expansion (Gedan et al., 2009). While human-induced degradation and loss have accelerated in recent decades (Deegan et al., 2012; Wang et al., 2014; Tian et al., 2016), the remaining tidal marshes are facing the additional global threat of accelerated sea level rise (SLR) caused by climate change (Spencer et al., 2016; Schuerch et al., 2018). In addition, the capacity of tidal marshes to adapt to SLR by sediment accretion and surface elevation gain can be compromised by decreasing sediment supply, for example due to upstream river damming and erosion control measures (Weston, 2014; Yang et al., 2020). Hence, efforts for conservation and restoration of tidal marshes are increasing throughout the world (Mossman et al., 2012; Liu et al., 2016; Zhao et al., 2016; Waltham et al., 2021), usually with primary goal to support and rehabilitate biodiversity (Armitage et al., 2007; Weinstein, 2007) and provide nursery habitat for commercially important fish and invertebrate species (Rozas and Minello, 2001). Furthermore, marsh restoration is increasingly motivated by its role for nature-based shoreline protection, as marshes attenuate waves, currents, and erosion and promote sediment accretion with SLR (Kirwan and Megonigal, 2013; Temmerman et al., 2013; Barbier, 2014; Kirwan et al., 2016; Zhu et al., 2020), and for nature-based mitigation of climate change impacts through carbon sequestration (Barbier et al., 2011; Rogers et al., 2019). The success of restoration designs largely depends on the resulting rates of marsh vegetation development and sediment accretion, as they control the timescales at which target habitats, effective shoreline protection, and carbon sequestration are reached. Besides, restoration designs must enable the development of marsh ecosystems that are resilient to modern threats such as SLR and decreasing sediment supply. Yet predicting actual rates of vegetation development and sediment accretion in establishing marshes, at timescales ranging from years to decades, remains an important challenge to this day (Fagherazzi et al., 2012; Mossman et al., 2012; Wiberg et al., 2020; Fagherazzi et al., 2020; Törnqvist et al., 2021).
Managed realignment, which consists in shifting the line of coastal defense structures landward of their existing position, can create space for tidal-marsh restoration or creation. This practice has grown in popularity over the last 2 decades (French, 2006; Turner et al., 2007), especially in the context of coastal squeeze and the landward movement of the mean low water mark due to SLR and storms (Doody, 2013). Practically, a second line of defense is built landwards; then the first one is breached. The number and size of breaches are important design choices (Hood, 2014, 2015) and vary greatly between projects (e.g., Friess et al., 2014; Dale et al., 2017). As breaches become the inlets of the restored marshes, they have an important control on water and sediment volumes entering and leaving the system during each tidal cycle and hence on sediment accretion rates (Oosterlee et al., 2020). Other important design measures may involve excavating an initial channel network and treating soil conditions to facilitate soil drainage (O'Brien and Zedler, 2006), manually planting vegetation tussocks to ensure vegetation encroachment (Staver et al., 2020), or building hydraulic structures to control the tidal range and create optimal ecological conditions for vegetation development (Maris et al., 2007; Oosterlee et al., 2018). These design choices are mainly driven by restoration objectives and local environmental conditions. Yet there is high uncertainty in how restored tidal marshes continue to develop over time. For example, several studies indicate that restored sites underperform in terms of biodiversity (Wolters et al., 2005; Mossman et al., 2012), topographic diversity (Lawrence et al., 2018), groundwater dynamics (Tempest et al., 2015; Van Putte et al., 2020), and biogeochemical functioning, including carbon sequestration (Santín et al., 2009; Suir et al., 2019) when compared to their natural counterparts. These outcomes can potentially hamper marsh ecosystem functions and the initial restoration objectives.
The rate at which tidal marshes develop in restoration projects is highly uncertain. In some cases, sediment accretion rates determine whether restored tidal marshes can keep pace with local rates of SLR (Kirwan et al., 2010; Vandenbruwaene et al., 2011a; Webb et al., 2013; Kirwan et al., 2016). The establishment rate of pioneer vegetation and the succession towards climax vegetation may depend on small windows of opportunity that are very difficult to predict (Chambers et al., 2003; Hu et al., 2015; Cao et al., 2018). Furthermore, the rate of development is at the center of the tension between public perception and restoration objectives. The public is often very critical towards marsh restoration by managed realignment, as it implies the loss of valuable land, laboriously reclaimed by previous generations (Temmerman et al., 2013). On the one hand, fast development allows quickly reaching target habitats, which may support a positive public perception, but involves the risk of fast development towards a monotone climax ecosystem state. On the other hand, slow development (e.g., including bare mudflats) increases the risk of negative public perception in the first years but may lead to long-term persistence of high habitat diversity with different stages of succession. All these examples illustrate the need for modeling tools that can predict how fast restored tidal marshes develop and how development rates can be steered by restoration design.
Numerical models of tidal-marsh development need to be able to simulate vegetation and landform evolution through feedbacks between hydrodynamics,
sediment transport, and vegetation dynamics. For example, sediment accretion rates determine where and when the critical elevation, above which
vegetation can grow in the intertidal zone, is reached (Bertness and Ellison, 1987; Balke et al., 2016; Bouma et al., 2016). The encroachment of
vegetation in turn enhances sediment accretion through tidal flow deceleration, direct sediment trapping, and organic matter accumulation
(Vandenbruwaene et al., 2011b; Baustian et al., 2012; Fagherazzi et al., 2012). Sediment accretion processes are intrinsically linked to the
availability of sediments at the marsh edge (Weston, 2014; Liu et al., 2021) as well as their tidal distribution along complex channel networks and
towards the vegetated platforms (Marani et al., 2003; Temmerman et al., 2005). At the same time, vegetation is known to control the morphology of
channel networks and their efficiency to deliver sediments into the marsh interior (Kearney and Fagherazzi, 2016; Schwarz et al., 2018). So far,
spatially explicit models that combine detailed hydro-morphodynamics with complex vegetation dynamics remain uncommon, and their applicability is
limited to relatively small domains (on the order of 1
In this paper, we present a biogeomorphic model application to a specific tidal-marsh restoration project by managed realignment, accounting for
relevant fine-scale flow–vegetation interactions (less than 1
Schematic representation of the multiscale biogeomorphic coupling.
We have developed the biogeomorphic modeling framework Demeter to simulate explicitly, in an intertidal landscape, the feedbacks between (i) tidal
hydrodynamics, (ii) sediment erosion, transport, deposition, and resulting bed level changes, and (iii) vegetation establishment, expansion, and
die-off for multiple dominant vegetation types. We adopt a multiscale approach (Fig. 1a), in which the hydro-morphodynamics are computed at a grid
resolution of 5
Demeter is in principle compatible with any process-based hydro-morphodynamic model. Here we use the finite element solver suite Telemac (version 7.3.0) and more specifically its modules Telemac-2D for hydrodynamics and Sisyphe for sediment transport and morphodynamics. Telemac-2D simulates water level fluctuations and flow velocities by solving the depth-averaged shallow-water equations in a two-dimensional horizontal framework (Hervouet, 2007). The vegetation resistance force is modeled as the drag force on a random array of potentially submerged rigid cylinders with uniform properties per species (Baptist et al., 2007) and calibrated against flume measurements (Vandenbruwaene et al., 2011b; Bouma et al., 2013; Gourgue et al., 2021a). Sisyphe simulates the transport of cohesive sediments by solving the depth-averaged advection–diffusion equation as well as bed level changes through sediment erosion (Partheniades, 1965) and deposition (Einstein and Krone, 1962).
Demeter includes a cellular automaton that is used here for the vegetation dynamics. A cellular automaton consists of a regular grid of cells, each one with a finite number of states (here, vegetation species). Cells can change their state in discrete time steps, depending on their neighborhood state and a set of simple stochastic transition rules (Balzter et al., 1998). The cellular automaton here simulates the fate of three generic species representative of pioneer, middle-marsh, and high-marsh plants encountered near the study site. Transition rules (i.e., for vegetation establishment, lateral expansion, species competition, and vegetation die-off) are determined based on field observations and depend on environmental stressors provided by Telemac (i.e., hydroperiod, bed shear stress, and bed level changes).
Based on observations in the vicinity of the study site, expected vegetation species that are representative of pioneer, middle-marsh, and high-marsh
vegetation are, respectively,
During one sequence of hydro-morphodynamics, Telemac keeps track of different environmental stressors (i.e., hydroperiod, bed shear stress, and bed level changes), which are then used to regulate the transition rules for vegetation establishment, lateral expansion, species competition, and vegetation die-off in the vegetation module. During one sequence of the vegetation module, the cellular automaton updates the distribution of vegetation, which is then used to evaluate the vegetation resistance force in the hydro-morphodynamic module. See Sects. S1.3 and S1.4 for more details.
The computation sequence is schematized in Fig. 1b. We use the morphological acceleration method to decouple hydrodynamic and morphodynamic timescales
(Lesser et al., 2004; Roelvink, 2006). The hydrodynamics are computed with time steps of 1.5
Overview of the study site location within northwestern Europe
Hedwige–Prosper Polder (Fig. 2) is an agricultural area of 4.65
Local environmental conditions are a determinant for the development of restored ecosystems (Liu et al., 2021). The Scheldt Estuary, here defined as the tidal part of the Scheldt River, is a semi-diurnal macrotidal estuary extending over 160
Model scenarios.
We investigate the resilience of the restored tidal marsh to human-induced climate and environmental changes by considering different relative SLR
rates and different SSC at the seaward boundary (Table 1). If our model can account for changes in MSL (Sect. 2.1), changes in MHWL are more relevant
for the biogeomorphology of tidal marshes, as the intertidal elevation relative to MHWL determines the tidal inundation regime, hence affecting
sediment accretion rates (Temmerman et al., 2004) and vegetation growth (Balke et al., 2016). Therefore, for the reference model scenario, we consider an SLR rate corresponding to the average rate of MHWL rise observed in the Scheldt Estuary over the last century, that is, 6
Human activities can also potentially disturb the estuarine sediment dynamics. For example, river damming and erosion control measures in the upstream
river catchment can potentially reduce the sediment supply to the estuary (Syvitski et al., 2009; Yang et al., 2020), while increasing tidal intrusion
can potentially increase the SSC (Winterwerp et al., 2013). As sediment supply is critical for marsh development (Hopkinson et al., 2018; Liu et al.,
2021), we consider three scenarios with different SSC at the seaward boundary, that is, the reference scenario with an SSC of 63
We also investigate the impact of restoration design on the rates of spatial vegetation and landform development. In particular, the width of the dike
breaches (i.e., the inlets of the restored marsh) is an important design choice in managed realignment projects. Here, we simulate four different
versions of the small inlet (Fig. 2d–g, Table 1). We first consider the reference design (Fig. 2d), which simply consists of a 50
In the reference model scenario, vegetation establishes itself randomly following different colonization strategies in areas where environmental stressors allow for it. Pioneer marsh vegetation establishes itself homogeneously with a relatively high probability of establishment but with no possibility of expanding laterally. Middle- and high-marsh vegetation establishes itself patchily with a relatively low probability of establishment but with the possibility of expanding laterally to form growing patches (Sect. 2.1.2). This is the expected behavior supported by field observations for the three selected species (Sect. S1.5.2). To illustrate the impact of the vegetation dynamics on the biogeomorphic feedbacks and the model results, we also consider six variants of the reference model scenario (Table S1 in the Supplement).
We have developed a model to predict the biogeomorphic evolution of a restored tidal marsh in an area that is still embanked. Due to the lack of in situ data to validate our modeling approach, we evaluate its overall performance against observations in nearby intertidal mudflats and marshes. This is done for the reference model scenario (Sect. 2.3 and Table 1). Observations on elevation change (Sect. 2.4.1) and vegetation development (Sect. 2.4.2) are used for qualitative model calibration, based on model sensitivity to a selection of sediment parameters (i.e., critical shear stress for erosion, settling velocity, dry bulk density, and incoming SSC) and vegetation parameters (i.e., establishment probability and patch expansion rate). Observations on channel network characteristics (Sect. 2.4.3) are used for qualitative model validation. In some cases, we calculate linear regressions from both model results and observations, and we perform an analysis of covariance (ANCOVA) to determine whether they are significantly different from each other (Sect. 2.4.4).
Sediment accretion processes on vegetated platforms are crucial for marsh resilience against SLR (Kirwan et al., 2016). Based on digital elevation maps derived from historical topographic surveys in the adjacent marshes of the Drowned Land of Saeftinghe (Fig. 2c) between 1931 and 1963 (Wang and Temmerman, 2013), we have developed an empirical relationship between mean elevation change on vegetated platforms and mean high-water depth. Here, we develop a similar relationship based on model results in the restored tidal marsh, using the same variables over the same time interval (i.e., between years 18 and 50 after de-embankment), and we compare it with the empirical relationship derived from observations. See Sect. S2 for more details.
The encroachment of vegetation on intertidal platforms impacts sedimentation in tidal marshes and hence their geomorphic development (Mudd et al., 2010). Here, we compare our model results with the observed rate of spatial expansion of the vegetation cover in the adjacent restored marshes of Paardenschor (Fig. 2c), from its de-embankment in 2004 until 2017. See Sect. S3 for more details.
Channel networks control the flow of water and sediments in tidal marshes, and their evolution interacts with the biogeomorphic development of the surrounding intertidal platforms (D'Alpaos et al., 2007; Kearney and Fagherazzi, 2016). Here, we compare various geometric properties of the simulated tidal channels with observations in the adjacent marshes of the Drowned Land of Saeftinghe (Fig. 2c – Vandenbruwaene et al., 2013, 2015). To that end, we have developed a quasi-automatic methodology to extract tidal channel networks and related geometric properties from model results. More specifically, we compute the probability distribution of unchanneled flow length (i.e., the shortest distance to a channel bank) as a measure of channel density (Tucker et al., 2001). The mean unchanneled flow length is calculated as the slope of the linear portion of the probability distribution when plotted on semi-log axes (Marani et al., 2003; Chirol et al., 2018). Along the channel network skeleton (i.e., the channel centerlines – Fagherazzi et al., 1999), we compute the watershed area, the upstream mainstream channel length (i.e., the longest upstream channel within the watershed), the mean overmarsh tidal prism (i.e., the mean high-tide water volume within the watershed for all tides overtopping the surrounding platform – Vandenbruwaene et al., 2013, 2015) and the channel cross-sectional dimensions (i.e., channel width, channel depth, and channel cross-section area). We also verify the applicability of Hack's law, an empirical power relationship that links watershed area and mainstream channel length (Rigon et al., 1996). See Sect. S4 for more details.
In several cases, we calculate linear regressions from both model results and observations. First, we split the data into 10 sub-samples of equal
size, based on the
Elevation data are either expressed with reference to the Normaal Amsterdams Peil (NAP), the official vertical datum in the Netherlands, which represents approximately the MSL at the Dutch coast, or the local MHWL.
The mean platform elevation over a certain area (i.e., the Northern Basin, the Southern Basin, or both combined) is calculated excluding channels and supratidal areas (i.e., the highest parts of the dikes and constructed bird breeding islands). The vegetation cover over a certain area is calculated as the proportion of vegetated cells, excluding supratidal areas.
The ebbing time is a proxy for the drainage efficiency of an inlet and is calculated as the time needed to drain 95 % of the total ebb volume. The ebb volume is the instantaneous volume of water flowing through an inlet since high tide during the ebb phase (i.e., from high to low tide in the estuary). The total ebb volume is the ebb volume at the end of the ebb phase (i.e., at low tide). The relative ebb volume is the instantaneous ebb volume divided by the total ebb volume.
Reference model scenario (no. 1). Bed elevation
Reference model scenario (no. 1).
The results of the reference model scenario indicate that the system is predominantly depositional (Fig. 3f–j), with only localized erosion of up to
about 10
In the first 10 to 20 years, most sedimentation occurs at the back of the Northern Basin and almost everywhere in the Southern Basin (Fig. 3a, b and f, g), where pioneer vegetation and then middle-marsh vegetation start to encroach whenever the elevation and hence the hydroperiod become suitable (Fig. 3k and l). The middle-marsh vegetation's ability to colonize its surroundings via clonal expansion (Sect. 2.1) is especially visible in years 20 and 30 (Fig. 3l and m). In general, middle-marsh and high-marsh vegetation follows similar colonization strategies but with a 10- to 20-year delay (Fig. 3l and o).
In the first 20 to 30 years, some of the initially excavated channels are predicted to fill up (Fig. 3f–h) or even to disappear over time (Fig. 3c–e), especially those oriented perpendicular to the tidal flow propagation. Meanwhile, newly formed channels are primarily created by differential deposition (Fig. 3g and h), although erosion starts to occur in a second phase when they become narrower (Fig. 3h–j). In general, vegetation does not encroach on tidal channels before they fill up because the hydroperiod and the shear stress are too high.
Overall, the presence of vegetation slightly increases the rate of platform accretion in the Northern Basin, although the speed of colonization has nearly no influence on the mean platform elevation 50 years after de-embankment (Fig. S1a in the Supplement). In the Southern Basin, neither the presence of vegetation nor the speed of colonization seems to affect sediment accretion on the platforms (Fig. S1b), which suggests that the hydrodynamics are predominant in that part of the restored marsh. Locally, the vegetation dynamics can have remarkable geomorphic effects, such as the maintenance or disappearance of pre-excavated channels, whether we consider no vegetation, the reference vegetation dynamics, or instantaneous colonization (Fig. S2). In general, vegetation input parameters have a rather limited impact on the long-term morphodynamics (Fig. S3).
Our results for the reference model scenario are in good agreement with the empirical relationship between mean elevation change on vegetated
platforms and mean high-water depth, which is obtained from observations in a tidal marsh close to the study site (Fig. 4a). The linear regressions
obtained from model results and observations are statistically equal, as both their slopes (
Reference model scenario (no. 1). Channel geometric properties 50 years after de-embankment (blue) compared to observations in an established marsh close to the study site (black). Probability distribution of the unchanneled flow length
The predicted channel network 50 years after de-embankment is slightly less dense, as compared with observations in a nearby tidal marsh (Fig. 5a),
with mean unchanneled flow lengths of 26.0
Sea level rise rate
Our results indicate that the restored tidal marsh is resilient to realistic ranges of SLR rate and SSC at the seaward boundary (Fig. 6). Indeed, for every model scenario considered, the intertidal platforms accrete faster than SLR (Fig. 6a and b) and the restored area can sustain a growing vegetation cover (Fig. 6c and d).
In Fig. 6a and b, we show the mean platform elevation with respect to MHWL, which increases over time due to SLR, rather than a fixed level like NAP. This is especially relevant when comparing the system response to different SLR rate scenarios. In terms of absolute platform elevation (i.e., with respect to a fixed level like NAP), increasing rates of SLR lead to faster platform accretions. However, the rate at which intertidal platforms catch up with MHWL decreases with increasing rates of SLR (Fig. 6a). This is especially relevant in terms of marsh resilience, which is defined as the ability to keep pace with SLR. Platform elevation (Fig. 6a and b) is also related to vegetation development (Fig. 6c and d), which is driven by hydroperiod rather than absolute elevation. That is why we choose to present the mean platform elevation with respect to the MHWL.
By the end of the 50-year simulated period, the system has not reached an equilibrium where both mean platform elevation and vegetation cover converge to a stable value. Nevertheless, it can be extrapolated from our results that equilibrium will occur more slowly for an increasing rate of SLR (Fig. 6a and c) and decreasing sediment supply (Fig. 6b and d). Interestingly, within the range of considered parameter values, the rate of biogeomorphic development seems more sensitive to suspended sediment availability (Fig. 6b and d) than SLR rate (Fig. 6a and c).
To investigate the impact of inlet width, we first analyze separately how the large-inlet Northern Basin and the small-inlet Southern Basin (Fig. 2c) behave for the reference scenario. We then compare model scenarios with increasing inlet widths in the Southern Basin.
Reference model scenario (no. 1). Evolution of the mean platform elevation with respect to the mean high-water level (MHWL)
Reference model scenario (no. 1). Water depth (color maps) and flow velocity (gray arrows) every 50
Our results indicate that, for the reference design, the Southern Basin develops faster than the Northern Basin (Fig. 7a and b). During the flood phase (Fig. 8a–c), water and suspended sediments fill the Southern Basin through the small inlet but also from the Northern Basin. Conversely, in the first part of the ebb phase (Fig. 8d and e), water and suspended sediments leave the Southern Basin both through the small inlet and towards the Northern Basin. However, in the second part of the ebb phase (Fig. 8f), the water surface elevation in the Southern Basin drops below the Northern Basin platforms, so that water and suspended sediments can only flow towards the small inlet. Meanwhile, the small inlet slows down the evacuation of the remaining water, giving more time to suspended sediments to settle in the Southern Basin.
Reference model scenario (no. 1). Water volume
This interpretation is confirmed by an analysis of the water volume and sediment mass budgets during a spring tide in the first year after de-embankment (Fig. 9). At the boundary between the Northern and Southern basins, there is 40 % more water entering the Northern Basin during the flood phase than leaving it during the ebb phase (Fig. 9a, pink). Consequently, the small-inlet ebb discharge is 40 % higher than its flood discharge (Fig. 9a, blue). Besides, it takes 40 more minutes to flush the water through the small inlet than through the large inlet (Fig. 7c), which gives 30 % more time for suspended sediments to settle in the Southern Basin after high tide. This leads to 60 % more sedimentation in the Southern Basin than in the Northern Basin (Fig. 9b, black). The same pattern can be observed, albeit to a lesser degree, for intermediate and neap tides as well as at later stages of marsh development.
Inlet design model scenarios (i.e., reference design and three alternative designs with small-inlet widths of 50, 100, and 200
Inlet design model scenarios (i.e., reference design
Our argument that faster biogeomorphic development in the Southern Basin is due to enhanced sediment trapping related to its small inlet size is further supported by the three alternative design scenarios (Fig. 2e–g). Our results do indeed demonstrate that widening the small inlet slows down the biogeomorphic development in the Southern Basin (Fig. 10b and d). It has however the opposite effect in the Northern Basin, where widening the small inlet leads to faster biogeomorphic development (Fig. 10a and c). At the landscape scale, widening the small inlet leads to the formation of more, larger channels that originate from the small inlet and expand into the Northern Basin (Fig. 11, ellipses). This contributes to the supply of more sediments into the Northern Basin. Interestingly, while the width of the small inlet clearly affects the spatial distribution of sedimentation and vegetation development in the restored marsh, it barely impacts the general biogeomorphic trend when the Northern and Southern basins are considered combined (Fig. S4). Overall, it is a zero-sum game where the Northern and Southern basins act like communicating vessels.
The ebbing time remains very close to an average value of 3
The increasing demand for the creation and restoration of resilient tidal marshes calls for the development of new modeling tools. These models must be able to assess whether rates of biogeomorphic development will achieve restoration objectives and how these rates can eventually be steered by restoration design. In this paper, we present a novel biogeomorphic model application to a tidal-marsh restoration project in a macrotidal estuary. We find that the ability of the restored marsh to sustain future rates of SLR is more sensitive to suspended sediment availability than to rates of SLR (Fig. 6). We also demonstrate that inlet design can steer rates and spatial patterns of biogeomorphic development (Figs. 10 and 11) and as such can be used to achieve specific restoration objectives.
One of the main objectives in every tidal-marsh restoration project is for intertidal platforms to build vertically faster than SLR, allowing us to develop and maintain a vegetation cover. In that context, local environmental conditions play a crucial role. In the case studied here, our model
predicts that, for the first 50 years, the restored tidal marsh can keep pace with realistic rates of SLR and that its resilience is more sensitive
to suspended sediment availability (Fig. 6). These findings are in line with previous studies on marsh adaptability to SLR. For example, an ensemble
model study indicates that tidal marshes with similar sediment input and tidal range as our study site can cope with SLR rates of up to
70
A second key question that is typically raised when planning for tidal-marsh restoration is how fast the ecosystem and its different habitat zones develop (Yando et al., 2019), which is for a large part driven by the rates of sediment accretion, pioneer vegetation establishment, and succession. Here, the desired rates depend on the restoration objectives. High rates of sediment accretion and vegetation development allow us to quickly reach certain restoration objectives, such as different aspects of nature-based shoreline protection. For example, high-lying and densely vegetated marshes are most effective for wave attenuation (Möller, 2006; Schoutens et al., 2020; Willemsen et al., 2020) and reduction of shoreline erosion (Möller et al., 2014; Francalanci et al., 2013; Wang et al., 2017; Schoutens et al., 2019) and dike breaching hazards (Zhu et al., 2020). However, when tidal marshes are created along estuaries or deltas to attenuate extreme high tides and storm surges (Smolders et al., 2015; Stark et al., 2017; Huguet et al., 2018; Smolders et al., 2020), lower accretion rates are preferred to maintain higher water storage capacity in the restored tidal marshes. When the objective is to restore intertidal habitats and meet specific biodiversity goals (Hinkle and Mitsch, 2005; Chang et al., 2016), it may be favorable for accretion rates to be high so that vegetation can develop fast but also not too fast so that a diversity of habitats can persist over time, including tidal channels, mud flats, pioneer marsh, and higher-marsh vegetation while avoiding a rapid succession to climax species. In the case studied here, our model predicts that the reference restoration design can achieve the objective of intertidal biodiversity rehabilitation. Indeed, because of the relatively slow accretion rates in the Northern Basin (Fig. 7), the restored tidal marsh still features the entire range of intertidal habitats after 50 years (Fig. 3).
The examples above illustrate the need to identify restoration design options that can steer rates of sediment accretion and vegetation development in line with restoration objectives. In this study, we focus on the impact of one specific design option: the inlet width (Fig. 2). Our model predicts that, for the setting studied here, higher differences between the two inlet widths lead to more contrasting sedimentation and vegetation patterns in the two basins (Fig. 10). This has two positive outcomes. On the one hand, high accretion rates in the Southern Basin bring fast vegetation development there and potentially positive public perception for the restoration project. On the other hand, lower accretion rates in the Northern Basin allow for long-term persistence of habitat diversity, which is an important objective in this project. Other important design options, such as the excavation of a channel network (Williams et al., 2002; Wallace et al., 2005; Hood, 2014), the manual planting of vegetation (O'Brien and Zedler, 2006; Staver et al., 2020), the infilling or lowering of areas, or the creation of a landward slope, were beyond the scope of this study. However, new fundamental insight on the impact of such design options is crucial and should be investigated in the future. Our novel biogeomorphic model is made available for the scientific community with this in mind (see “Code and data availability” section).
An important challenge for tidal-marsh biogeomorphic models is that rates of sediment accretion and vegetation development are highly uncertain (Wolters et al., 2005; French, 2006; Mossman at al., 2012; Spencer and Harvey, 2012). For example, exposure to wind waves and resulting sediment resuspension processes (Leonardi et al., 2016) can considerably lower accretion rates and potentially prevent any vegetation from establishing itself (French et al., 2000). In contrast, improper inlet design can lead to very high accretion rates of poorly consolidated sediments that also limit vegetation establishment (Oosterlee et al., 2020). Other important sources of uncertainty for vegetation development include biotic control such as benthic animals grazing on establishing plants (Paramor and Hughes, 2005; Silliman et al., 2005) and plant stress related to extreme events such as droughts or hurricanes (Howes et al., 2010). At this stage, our model is not equipped to evaluate these uncertainties.
However, to our knowledge, this paper presents the first application of a tidal-marsh biogeomorphic model accounting for relevant fine-scale
interactions (less than 1
Studies that evaluate the performance of a tidal-marsh biogeomorphic model against field observations of marsh development over relevant
spatiotemporal scales (several
In this paper, we present a biogeomorphic model application to a specific tidal-marsh restoration project by managed realignment at the Belgian–Dutch border along the macrotidal Scheldt Estuary. Our results indicate that the restored tidal marsh can keep pace with realistic rates of SLR and that its resilience is more sensitive to suspended sediment availability, in agreement with current scientific knowledge. Our results further demonstrate that restoration design options can steer the biogeomorphic development of restored tidal marshes and as such can be used as means to achieve specific restoration objectives.
Predicting the rate of restored tidal-marsh development is very important, yet highly uncertain. Our novel biogeomorphic model, whose performance is here evaluated against observations in tidal systems close to the study site, is an attempt to reduce that uncertainty. Our promising results call for the evaluation of additional restoration design options.
The biogeomorphic model Demeter is available at
The supplement related to this article is available online at:
OG, JvB, CS, TJB, JvdK, and ST conceptualized the model and the study. OG implemented the model. OG and JvB evaluated the model performance and analyzed the results, with contribution from JPB and under the supervision of TJB, JvdK, and ST. WV and JV provided resources to run the model and evaluate its performance. OG wrote the paper under the supervision of SF and ST and with contribution from all co-authors
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank the Associate Editor Claire Masteller and two anonymous referees for their constructive remarks that helped improve the paper. The resources and services used in this work were provided by the VSC (Flemish Supercomputer Center), funded by the Research Foundation – Flanders (FWO) and the Flemish Government.
This project has received funding from the Vlaams–Nederlandse Scheldecommissie (VNSC), the European Union's Horizon 2020 research and innovation program (Marie Skłodowska-Curie Actions – global postdoctoral fellowship – grant no. 798222), and the Research Foundation – Flanders (FWO – fundamental research project – grant no. G060018N). Sergio Fagherazzi was partly funded by the USA National Science Foundation awards 1637630 (PIE LTER) and 1832221 (VCR LTER).
This paper was edited by Claire Masteller and reviewed by two anonymous referees.