This flume experiment was conducted in the Mesodrome flume facility at the University of Antwerp (Belgium) ( Figure 2A ). The flume consists of a 10 m long, 2.0 m wide and up to 1.5 m deep test section and has a maximum pump capacity of 0.6 m³ s^-1. To generate very high flow velocities the width of the test section was reduced to 0.8 m. Within the test section, 8 monoliths were placed to create a marsh of 0.8 m wide and 9.6 m long ( Figure 2B ).

The monoliths were excavated from tidal marshes in the Scheldt estuary ( Figures 3A, B , 51.35 N, 4.23 E) as sediment blocks of 0.8 m by 1.2 m in surface area and 0.4 m depth, with vegetation growing on top. The monoliths were excavated on January 20 and 21, 2021, from brackish tidal marshes dominated by a mature Phragmites australis (Cav.) Trin. Ex Steud. Vegetation ( Figures 3C, D ) which can grow up to 4 meter high ( Supplementary Figure 1 ).

Extraction was done by digging a pit, as such creating a vertical cliff in the sediment; then pushing a metal plate horizontally into the cliff at a depth of 0.4 m under the horizontal sediment surface ( Supplementary Figure 2A ); next a rectangular mold of 0.8 by 1.2 m in surface area and 0.4 m high was placed on top of the sediment surface and pushed down gently until it made contact with the metal plate ( Supplementary Figure 2B ). The marsh monolith, contained within a “box” created by the plate and the mold, was lifted and placed on a pallet covered with a thin horizontal, perforated multiplex board (i.e. to support the sediment block, to prevent cracks and to allow a little bit of drainage during the flume experiments). The mold was removed and vertical multiplex boards were attached around the monolith for transport by a truck ( Figure 3D and Supplementary Figure 2C ). After placing the monoliths in the flume with a crane, the protective vertical boards were removed from around the monoliths and they were positioned along the 10 m test section ( Supplementary Figure 2D ).

Aboveground biomass of the first 5.6 meters of the monoliths (at the leading edge of the test section) was removed to create a zone without vegetation for the incoming flow before that entered the zone with vegetation remaining along a length of 4 m at the end of the test section ( Figure 2B ). The remaining shoots were cut at a height just below the wooden beams crossing the flume ( Figure 2A ) to exclude interference of bending of the shoots with the beams. The beams were needed to keep the setup in place and withstand the hydrodynamic forces. As such the remaining vegetation stems were max. 1 m high. Small gaps in between the different monoliths and at the side edges were filled with sediment to ensure that the sediment bed was continuous across the flume test section. Before the monoliths were exposed to flow velocities, the flume was filled up with water to the sediment surface to let the sediment in the monoliths settle for two days. A first set of 8 monoliths was used as a pilot experiment to find out the desired settings (i.e. combination of water level and flume pump rate) to create maximum possible flow velocities with this flume setup. Next, the second set of 8 monoliths was installed in the flume to have undisturbed monoliths before starting the measurements of tidal marsh stability.

The experiments were conducted as six separate runs of 2 hours. Flow velocity and water depths remained constant during each run, but progressively increased from 1.00 to 1.75 ms^-1 and from 0.15 to 0.35 m with each new run ( Table 1 ). Flow conditions were defined by the Reynolds number (Re):

with v_in as depth averaged flow velocity (ms^-1), R as hydraulic radius which is defined as the cross-sectional area, A (i.e. for rectangular flumes: A = w.d with w being the flume width which was 0.8 m), divided by the wet perimeter of the flow (i.e. for rectangular flumes: 2d + w) and v being the kinematic viscosity (10^-6 m²s^-1 for water); and the Froude number (Fr):

with g as the gravitational acceleration (9.81 ms^-2) and D as the hydraulic depth which is defined as A/w, which in a rectangular flume is equal to the water depth, d.

The effect of aboveground vegetation cover on the sediment stability was tested by starting with two runs with the original vegetation cover present over 4 m of length of the test section (run 1 and 2), followed by consecutively manual removal (clipping) of 2 m of vegetation (run 3 and 4) and ending without vegetation cover (run 5 and 6). Flow velocities were measured every run with an electromagnetic flow meter (EMF, Valeport model 801, Totnes, UK) in the middle of the flume width and along a vertical depth gradient with an interval of 5 cm. The flow was measured 4.0 m and 0.8 m before the end of the test section ( Figure 2B ). Flume-wall effects on the flow were regularly checked by expanding the measurement of flow velocities over a cross-sectional grid. At the same positions, water depth was measured to calculate the slope of the water surface.

Site-specific sediment composition was quantified on six surficial sediment samples after removing the top layer of plant litter. Samples from the sediment bed were taken with a Kopecky ring (5.0 cm in diameter and 5.1 cm high and five replicates) and used to determine dry bulk density (after drying at 70°C for 72 h). Next, six mixed scrape samples of the top 2 cm were used to perform volumetric grain size analyses with a Mastersizer 2000 (Malvern) based on laser diffraction after a combined H₂O₂ and HCl treatment to remove organic compounds and disperse aggregates. Organic matter content was determined with the loss on ignition method, i.e., by ashing the samples for 4 hours at 550°C in a muffle furnace (Heiri et al., 2001). Shear strength was estimated based on four replicates with a pocket shear vane tester (Eijkelkamp, NL) for the surface sediment and a field inspection shear vane tester (Eijkelkamp, NL) at 10 cm depth. Penetration resistance of the sediment was measured on four replicates with a penetrologger (Eijkelkamp, NL) with a 1 cm depth interval, and an average of the upper 10 cm was calculated. All samples were taken in close approximation of the monolith extraction site, i.e. within 1-2 meters.

The marsh was covered with a homogenous P. australis vegetation in winter-state, i.e. the aboveground biomass consisted of dead, leafless stems, and leaf litter was lying on the sediment bed in between the standing stems (Supplementary Figure 3). The reed vegetation was characterized in the field in the same week as the monolith extraction (end of January 2021). Shoot densities were counted at three replicate 0.40 x 0.40 m square plots before all aboveground biomass was harvested and dried at 70°C for 72 h to quantify the aboveground biomass. Shoot lengths and basal shoot diameters were measured on 20 shoots, which were harvested to measure biomechanical properties, i.e. the flexural stiffness and Young’s modulus. For the latter, the basal 20 cm of the shoots were used to perform three-point bending tests with a universal testing device (Instron 5942, precision ± 0.5%). For more details on the methods to quantify the biomechanical properties we refer to Schoutens et al. (2021). Belowground biomass was quantified from five replicate sediment cores of 0.10 m diameter sampled up to 0.40 m depth (i.e. the same depth as the monoliths), which were sampled at the location of monolith extraction. The cores were frozen and cut into slices (0-2.5 cm; 2.5-5.0 cm; 5.0-10 cm; 10-20 cm; 20-30 cm; 30-40 cm). For each segment, the sediment was washed out and the remaining belowground biomass was dried (at 70°C for 72 h) and weighed.

Within the vegetated test section, 20 shoots were monitored during the first two runs (i.e. 4 meters of vegetation cover). In the third and fourth experimental run, 12 remaining shoots were monitored. In the fifth and sixth run, all vegetation was removed. The bending of the shoots in response to the high flow velocities was quantified in six categories indicating the shoot bending angle compared to the initial situation before the experimental runs. The categories ranged from shoots that did not suffer any damage or reconfiguration (< 5° bending angle) up to heavily bent shoots (> 35°) and broken shoots (i.e. flushed away or clearly broken shoots). The measurements were done at different moments in time, i.e. during the experimental run, directly after the run when the flow was stopped and after one (or three for run 4) day(s) of recovery.

Bed level changes (by erosion or sedimentation) were quantified by measuring the elevation of the sediment bed before and after every experimental run. Pin measurements [following a Sedimentation Erosion Bar, SEB, approach, see Nolte et al. (2013)] were performed along a grid over the 4 m long and 0.8 m wide vegetated part of the test section with a 5 cm interval in the direction parallel to the flume length and a 10 cm interval perpendicular to the flume length, revealing a total of 444 point measurements.

Tidal marsh monoliths were exposed to six consecutive runs of two hours each, with water depths ranging from 0.15 to 0.35 m and depth-averaged flow velocities ranging from approximately 1.00 to 1.75 ms^-1 ( Table 1 ). Flow velocities increased and the water surface slope decreased from runs 1 and 2 (with vegetation over 4 m of the flume length) to runs 3 and 4 (vegetation partially removed and remaining present over 2 m) and runs 5 and 6 (without vegetation cover) ( Table 1 ). The combined effect of the flume side walls and the rearrangement of the plant shoots (in response to the flow) towards the center of the flume caused variations of flow velocities over the width of the flume with a standard deviation of 0.1 – 0.2 ms^-1. Flow conditions during all runs were estimated to be sub-critical to nearly critical (estimated Froude numbers between 0.81-0.98) and highly turbulent (Reynolds numbers > 10⁵) ( Table 1 ). The sediment was characterized by a high silt fraction (~72%; 2-63 µm) and clay fraction (~17%; < 2 µm), around 20% of organic matter, and relatively high values of shear strength and penetration resistance ( Tables 2A, B ). P. australis vegetation in winter has a more modest aboveground biomass compared to the summer situation i.e., smaller, thinner and more flexible shoots ( Table 3 ). The majority of belowground biomass was found between 10 and 30 cm depth. At a depth below 30 cm, belowground biomass decreased ( Figure 4 ). The sediment bed is covered with a layer of litter (e.g. old leaves) under which a superficial network of fine roots can be found ( Figure 8 and Supplementary Figure 3 ).

Vegetation of P. australis in its winter state was able to cope with short-term very high flow velocities. The response of the reed vegetation and the sediment bed dynamics after an experimental run were the cumulative result of all previous experimental runs. During the two hour runs, the shoots were bent heavily in the direction of the flow, but recovered rapidly after the flow was stopped ( Figure 5 ). Even after four consecutive runs, total damage remained limited as less than 17% of the sampled shoots were broken and less than 17% were bent more than 35°.

Over the entire period of the six experimental runs (i.e., 12 hours cumulated exposure time), vertical erosion was limited to a median (± standard deviation) cumulative total erosion of 6.7 ± 2.4 mm only. Over individual runs (2 hours), the median erosion was maximum 2.5 ± 2.5 mm (for run 2) and less than 1.0 mm for all other runs. The first run removed part of the litter and organic debris that was initially covering the sediment surface ( Figure 6 and Supplementary Figure 3 ). In the sections where aboveground vegetation was removed, even more litter was removed in the run directly after removal (i.e. in the 4.0 – 2.0 m and 2.0 – 0.0 m distance from the end of the test section in run 3 and run 5 respectively). The presence or absence of standing P. australis shoots had no effect on the bed elevation changes (e.g. ANOVA comparing the vegetated section with the non-vegetated section in run 4: F _1,387 = 0.171, p = 0.68) and no systematic spatial patterns were found throughout the six runs ( Figure 6 ). Although there is a general trend of slight erosion, at some locations sediment accretion was also observed ( Figures 6 , 7 ). Apart from the general trend of limited erosion, outliers of several centimeters of erosion and deposition were observed throughout the entire experiment, as a result of translocated sediment aggregates ( Figure 7 ). After the 5^th and 6^th run, i.e. respectively after 10 h and 12 h of cumulative high flow velocities, first signs of uprooting appeared and revealed a shallow subsurface mat of fine roots ( Figure 8 ).

The results in this study suggest that the reduction of aboveground biomass in winter, reduces the experienced drag which then promotes the resistance of the vegetation against high flow velocities. We tested the winter state stability of P. australis vegetation, which is a typical dominant species in the high intertidal zone of brackish tidal marshes in NW European estuaries and in many other brackish and freshwater tidal estuaries worldwide (Srivastava et al., 2014). We found a high capacity of P. australis to withstand high flow velocities, as most of the aboveground stems (83%) did not break, of which 66.5% had a less than 25° bending angle at the end of all flume runs ( Figure 5 ). At first sight, this finding of high resistance of P. australis to high flow velocities may be contrasting with previous studies, showing that P. australis vegetation has a lower tolerance to strong hydrodynamic forces, in comparison to the pioneer marsh species that grow on lower intertidal elevations in the brackish parts of NW European estuaries (Coops et al., 1996; Asaeda et al., 2005). Yet, this apparent contradiction between our findings and previous studies may be explained by the following hypotheses. Firstly, these previous studies focused on the growth of P. australis under average hydrodynamic conditions during a whole growing season instead of short-term extremes in winter conditions. The extreme flow conditions generated on the tidal marsh platform close to a dike breach are expected to last for only a limited time period i.e. usually only one or two high tides that coincide with the storm surge event with a marsh inundation depth of several meters and inundation time of two to four hours per high tide (Stark et al., 2015; Smolders et al., 2020; Zhu et al., 2020a). Once the storm surge has passed, normal tidal conditions prevail again, with very shallow water depths (<0.3 m during high spring tides and even no flooding during neap tides) and weak flow velocities (<0.3 ms^-1) typically found for the high intertidal marshes investigated here (Bouma et al., 2005a; Temmerman et al., 2012). Hence, the capacity of P. australis shoots to cope with high flow velocities is only required for a limited time period in the order of several hours, while previous studies assessed the growth response of P. australis to hydrodynamic conditions over a whole growing season (many months) (Coops et al., 1996; Asaeda et al., 2005). Secondly, P. australis drops its leaves in winter, which might further contribute to its higher capacity to withstand high flow velocities, as compared to the previous studies on P. australis in summer condition. The vulnerability of tidal marsh vegetation to experience stress from hydrodynamic forces is known to be dependent on plant traits that generate high drag forces on the shoots, i.e. plants with high biomass, tall shoots, and stiff stems experience higher drag forces under given hydrodynamic conditions (Bouma et al., 2005b; Schoutens et al., 2020). It should be mentioned that to fit the flume setup of this experiment, P. australis shoots were shortened, which could have promoted the resistance during the flow and the recovery process after the flow i.e., shorter shoots experience less drag force during the flow and less downward force during recovery. Nevertheless, compared to values reported in literature for typical P. australis summer vegetation, the vegetation from which monoliths were extracted for this experiment had a lower shoot biomass, and stems were thinner and smaller when monolith excavation took place at the start of the winter season in January 2021 (Table 3). Moreover, compared to the stiff shoots reported for P. australis in summer ( Table 3 ), our results suggest that the remaining winter shoots have a higher flexibility. This might be because the stiffest shoots brake and get washed away and only the slightly more flexible shoots remain in winter, which is in line with studies that monitored shoot stiffness of P. australis over an entire season (Zhu et al., 2020b).

P. australis winter vegetation in front of a dike can handle short-term high flow velocities, i.e. a cumulative 12 hours in this experiment, without losing much of the shoot biomass (<17% of stems). This finding indicates that after a dike breach has been repaired, the marsh has a high chance to continue providing its shoreline protection function through vegetation-induced attenuation of waves, currents and erosion (Möller et al., 2014; Spencer et al., 2015; Carus et al., 2016; Schoutens et al., 2019; Sheng et al., 2021; Zhang et al., 2022). Furthermore, through survival of the marsh vegetation, the marsh can also sustain its capacity to accumulate suspended sediments that are supplied during regular tidal inundations (Temmerman et al., 2003; Silinski et al., 2016a) and as such, to build up elevation in balance with long-term sea level rise (Temmerman et al., 2004). Studies have demonstrated that tidal marshes in front of dikes, that grow vertically in balance with sea level rise, are very effective in sustaining their nature-based mitigation of waves, and reduction of wave loads on the dikes, under future scenarios of sea level rise (Vuik et al., 2019; Zhu et al., 2020a). Our findings indicate that dike breaching behind marshes, after breaches are repaired, will not compromise this long-term nature-based shoreline protection function of the marshes.

The sediment surface in this experiment was highly resistant against erosion by high flow velocities ( Figures 5 , 6 ). Apart from some outliers, the elevation changes ranged predominantly between – 5 and + 5 mm. Here we note that this range of elevation change is not much more than the measurement accuracy of the SEB method (1.5 mm) (van Wijnen and Bakker, 2001; Nolte et al., 2013). Studies on the sediment stability of tidal marshes against vertical erosion under storm surge conditions confirm the highly stable nature of tidal marsh sediments, both found in flume studies mimicking storm conditions (Möller et al., 2014; Spencer et al., 2016) and field assessments after storms (Pennings et al., 2021). Nevertheless, there are also field studies reporting considerable erosion of marshes after very severe storms, such as in freshwater marshes on the Mississippi deltaic plain after the severe 2005 hurricane season (Howes et al., 2010). Moreover, the generated flow velocities in our study go beyond the flow conditions that may be expected on a marsh during a storm surge (Bennett et al., 2020). Nevertheless, the sediment bed remained stable which indicates that the shear strength of the sediment in our experiment was higher than the exerted shear stress (De Smit et al., 2021). In the subsequent paragraphs, we discuss one by one a total of four possible explanations for the observed extremely high erosion resistance of the sediment bed.

Firstly, sediment composition is known to be a key determinant for the susceptibility to erosion of sediment surfaces (Lo et al., 2017; De Battisti et al., 2019; Evans et al., 2021). Thus our finding of high sediment bed resistance to vertical erosion is likely to also be determined by case-specific sediment properties. The grain size distribution of the mineral sediment fraction plays a major role. Erosion resistance of marsh sediments is known to increase with decreasing grain size and associated increasing cohesiveness (Christiansen et al., 2000; Feagin et al., 2009; Lo et al., 2017). Tidal marshes that are situated in front of a dike are often situated relatively high in the tidal frame, being sheltered from the incoming hydrodynamics at the shoreward marsh edge and experiencing relatively shallow flooding at high spring tides, allowing fine sediments to settle. This may explain the high proportions of small sediment fractions, i.e. silt and, to a lesser extent, clay, in the tidal marsh monoliths used in our experiment ( Table 2A ). The high silt and clay content promotes sediment cohesion and is most likely the first reason why our sediment had such a high resistance against erosion.

Secondly, high elevated, mature marshes in front of a dike, like in our case, are typically characterized by strongly consolidated sediments (Tempest et al., 2015), which increases the erosion resistance of the sediment surface (Watts et al., 2003). This is in line with our penetration resistance measurements that were comparable with values found in NW European salt marshes ranging around 200 kPa in the upper sediment layers (e.g. Are et al., 2002; van de Vijsel et al., 2020). Dry bulk densities in this study indicate a soil texture that is favorable for root growth and water drainage (Bradley and Morris, 1990). Although bulk densities are often higher in deeper soil layers, i.e. more compacted, bulk densities in the top few centimeters in this study where in line with values measured in other, natural marshes along NW European marshes, ranging between 0.50 – 0.65 g.cm^-3 (Crooks and Pye, 2000; Watts et al., 2003; Tempest et al., 2015; Schulte Ostermann et al., 2021).

Thirdly, organic matter content has a positive impact on the erosion resistance of minerogenic tidal marsh sediments and consists of small organic substances and larger belowground root biomass. Our measurements of Loss on Ignition combine these two components and revealed a relatively high fraction of organic matter in the sediments (20.03 ± 0.88%), compared to values in literature ranging between 6-20% for natural mature marshes (Crooks and Pye, 2000; Watts et al., 2003; Tempest et al., 2015; Gillen et al., 2021). In addition to clay particles, small organic substances increase the sediment cohesiveness by forming an adhesive between sediment particles, creating bigger sediment aggregates. Nevertheless, in marshes dominated by organic material (i.e. 80-90% of the sediment fraction), the sediment properties will be different with a lower bulk density and less consolidation (Brooks et al., 2021) which reduces the sediment stability and increases potential erosion processes (Chambers et al., 2019; Himmelstein et al., 2021).

Fourthly, the presence of vegetation can increase the stability of the sediment by (i) reducing hydrodynamic forces due to friction between the aboveground shoots and the moving water and (ii) by providing structural rigidity for sediments and aggregates through the belowground root system (Vannoppen et al., 2015; Cahoon et al., 2020). The sediment surface stability under the short-term high flow velocities in this study was not affected by the presence or absence of aboveground biomass ( Figures 6 , 7 ). Rapid removal of the dead organic litter and debris covering the sediment surface was observed after the first experimental runs. In natural marsh conditions, however, the larger scales (i.e., less boundary conditions and larger surface areas) might result in a redistribution of this organic matter rather than complete removal as observed in this flume experiment, hence providing a local shielding function for the sediment surface. Although logistical restrictions did not allow to install fresh, undisturbed monoliths for every single run, the elevation changes in run 2-6 might be more similar to the observations after run 1 in which there was a redistribution of organic matter rather than a removal. During the high flow velocities, P. australis shoots were completely bent over ( Figure 5 ), hence the friction with the water column was reduced. Although not directly tested, the high fraction of organic matter and the high portion of root biomass at the sediment surface might have had an important contribution to the sediment stability. That is, a high fraction of root biomass near the sediment surface ( Figure 4 ) might cause a decrease in bulk density by creating pores and voids between the sediment particles (Brain et al., 2012; Chen et al., 2012; Jafari et al., 2019). Root networks change the sediment characteristics both through the presence of dead and living roots which then function as a structural framework for the sediment aggregates, promote the formation of pores and enhance drainage capacity, hence increasing the erosion resistance of the sediment surface (Gyssels et al., 2005; Grabowski et al., 2011; Evans et al., 2021). Only in the two highest flow conditions (after 10-12 h of cumulative flow exposure), uprooting of fine roots covering the entire sediment surface was observed ( Figure 8 ). In addition, when roots get exposed at the sediment surface, they will cause local turbulence, which may result in scour features (Bouma et al., 2009; Schoutens et al., 2021). This might suggest that when high flow velocities would continue for a much longer period of time, the top layer of sediment may get damaged or removed, exposing the subsurface sediment layers with lower belowground biomass. Moreover, vertical variations in the shape of the roots, e.g. from dense fine roots in the upper sediment layers towards sparser thicker roots in deeper layers (Gillen et al., 2021) or reduced belowground biomass with increasing depth (Howes et al., 2010), can alter the shear strength too. Nevertheless, subsurface shear vane tests indicated highly stable sediments limiting the erosion risk ( Table 2A ), even when the top layer of sediment including the dense root network is gone.

Despite the fact that we simulated maximal, near to critical flow conditions in our flume experiment, with depth averaged flow velocities up to 1.75 ms^-1 ( Table 1 ), we recognize that the water depths in our experiment were, for practical reasons, limited to maximum 0.35 m. Under extreme storm surge conditions, when water depths on marshes can be as much as 1.5 to 2.5 m (e.g. Stark et al., 2015; Zhu et al., 2020a), it may be well possible that maximum flow velocities over a marsh nearby a dike breach reach up to several ms^-1. Therefore, the present flume experiment needs to be regarded as a first test, which reveals promising results on tidal marsh stability and motivates further testing under larger water depths and higher flow velocities.

Current findings are based on one specific tidal marsh characterized by fine, cohesive sediments with a high shear strength and a monospecific P. australis vegetation. Although we argue that there are marshes with similar characteristics, we recognize that many other types of vegetated marshes exist e.g., minerogenic and organogenic marshes. Although this study provides insights in the stability of tidal marshes under high flow velocities, further research would be needed to confirm our findings for a wider range of hydrodynamic conditions and variety of marsh types.

Further research should focus on increasing the water depth and the flow velocities to simulate more extreme storm surge conditions. Field experiments with in situ, controlled dike breaches could be an option, however they remain logistically challenging (Peeters et al., 2019, Peeters et al., 2015; Wu et al., 2011). Furthermore, such controlled dike breach experiments are usually conducted during calm or moderate weather conditions and during normal tidal conditions, when water depths and flow velocities over marshes are small compared to extreme storm surge conditions. Finding suitable test locations with low elevated marshes in front of a dike is challenging, as the opportunities for in situ dike breach experiments are very limited. Lowering the marsh platform or lowering the dike could be an option to simulate the water depths and flow velocities expected during a dike breach. Experiments in large and deep flume facilities could be an alternative to control the water depth and flow velocity without taking into account the natural tidal cycling.Apart from experiments with more extreme flow conditions, exploring the sediment stability of different types of vegetated marshes is advised. Several studies point towards the role of plant traits in stabilizing the sediment based on the structure of the root network (Howes et al., 2010; Jafari et al., 2019; Gillen et al., 2021). In this study, P. australis formed dense, shallow mats of roots which could benefit the sediment stability compared to species that only develop roots in the deeper sediment layers, suggesting that the type of roots and their structure might play an important role in the sediment stability (Feagin et al., 2009; De Battisti et al., 2019). Other studies emphasize the effect of sediment composition, such as the decreasing sediment stability in function of an increasing fraction of organic substances (Gillen et al., 2021) or an increasing fraction of coarser, sandy sediments (Schoutens et al., 2019; Evans et al., 2021). Moreover, climate induced environmental changes might alter the stability of tidal marshes, i.e. increased stress from inundation by sea level rise and increased hydrodynamic forces from storms could alter the sediment dynamics (Schuerch et al., 2018; Jiang et al., 2020) and the growth of tidal marsh vegetation (Kirwan and Guntenspergen, 2012). Therefore, testing the sediment stability and stabilizing effect of different vegetation types in different environmental settings (e.g. climate) will contribute to the generality of our findings and could improve our understanding of how the stability of vegetated marshes under extreme flow velocities could change in the future. In the first place, we advise to test vegetated marshes which typically occur in front of dikes, such as salt marsh species (e.g. Elymus sp. or Spartina sp.), freshwater marsh species and tropical systems such as mangroves.In the present study, we focus on a mature, high elevated marsh adjacent to the dike. However, the response to extreme flow velocities might be different in marshes with a more complex geomorphology created by features such as creeks (Symonds and Collins, 2004). Moreover, young, low-lying developing marshes are expected to have different, less stable sediment characteristics (Evans et al., 2021), which is especially the case when tidal marshes are (re)created, e.g. in managed realignment projects (Tempest et al., 2015; Van Putte et al., 2020). Implementations of nature-based shoreline protection by tidal marshes has increased over the past decades and will further increase in the future. Many projects involve new marsh development on low elevated, initially bare flats in front of the embankment, where sediment accretion might be relatively fast (Vandenbruwaene et al., 2011; Oosterlee et al., 2018), resulting in less consolidated sediments, low bulk densities, poor drainage and hence lower shear strengths (Watts et al., 2003; Van Putte et al., 2020). Experience from existing tidal marsh (re)creation projects suggests that it might take several years before marshes develop with sediments that have enough strength to withstand high shear stresses (Fearnley, 2008; Kadiri et al., 2011; Tempest et al., 2015; Sha et al., 2018) as can be expected on the marsh platform in front of a dike breach. In this respect, further research is needed on the rates of development of marsh sediment strength, as marshes develop from young pioneer marshes to older established marshes.

Our results indicate that conservation, restoration and creation of tidal marshes in front of engineered flood protection structures such as dikes, provides an extra natural ‘barrier’ that can remain stable under high flow velocities, and as such may reduce flood depths and damage in case of breaching of the engineered dike. Note that this extra barrier-function requires stable high marshes which need time to develop, but which have the capacity to adapt and develop with changing environmental conditions such as sea level rise. Hence, instead of constructing dikes directly adjacent to the mean low water level, thereby embanking and losing pre-existing tidal marshes and diminishing the chance of new seaward tidal marsh establishment, we stress the benefits of providing enough space for tidal marshes in front of dikes. In areas where coastal development by building of dikes is ongoing, such as in Tianjin (China), Jakarta (Indonesia), or Port Harcourt (Nigeria) (Martín-Antón et al., 2016; Sengupta et al., 2018; Zhang et al., 2020), land reclamation often leads to degradation of many of the ecosystem functions and can bring extra costs that are often not accounted for in the reclamation project (Wang et al., 2010; Tan et al., 2020). Therefore, preserving part of the tidal marshes in front of newly constructed dikes is strongly advised.

In cases where no tidal marshes are present, the creation of suitable conditions for marsh development is recommended (Van Loon-steensma and Slim, 2013). At locations with enough space shoreward of the dikes (e.g. along shallow bays or lagoons), tidal marsh establishment can be facilitated by creating shallow, sheltered conditions by sediment nourishment and building (temporal) barriers to reduce hydrodynamics and capture sediments (Hofstede, 2003; Dao et al., 2018; Vuik et al., 2019; Baptist et al., 2021; Hu et al., 2021). Furthermore, active planting of marsh plants may eventually further enhance the chances of tidal marsh establishment (Tagliapietra et al., 2018).

At locations where space is limited shoreward of dikes (e.g. embanked estuaries with narrow shipping channels), it is worth to explore areas where there is space for more landward building of a new dike and breaching of the existing dike, so-called managed coastal realignment. Managed realignment creates a sheltered environment, suitable for tidal marsh development between the breached and the new dike. This managed realignment strategy has been implemented in several places over the last decades (Esteves and Williams, 2017; van den Hoven et al., 2022), for instance in Belgium where a total of 2500 ha of tidal marshes will be (re)created by 2030 (Temmerman et al., 2013). In places where managed realignment is not a desirable option, one could think about double-dike systems with transitional polders (i.e., see Figure 7 in Zhu et al., 2020). This implies that after a high marsh has established, the land can be converted back to its former (often agricultural) function, while still offering the benefits of an elevated foreshore. However, as stated before, it is important to account for the time needed to develop stable marsh sediments. This means that if we think nature-based solutions are a promising option to cope with sea-level rise, we must plan well in advance in order to initiate the development of new marshes as part of any kind of nature-based flood protection program.