Facilitating Salt Marsh Restoration: The Importance of Event-Based Bed Level Dynamics and Seasonal Trends in Bed Level Change

Intertidal salt marshes provide a range of valuable ecosystem services which typically increase with marsh width. Understanding the drivers for salt marsh expansion versus retreat is thus key to managers. Previous research highlights the influence of short-term (daily/event) bed level dynamics on germination and establishment and subsequent vegetation presence. However, more recent literature suggests the importance of medium-term seasonal bed level dynamics on viable seed availability and subsequent vegetation presence. This study aims to assess event-based and seasonal bed level dynamics for vegetation presence in natural and semi-natural salt marshes and to provide generic thresholds for vegetation presence. To gain insight into bed level dynamics, data was used from autonomous Optical and Acoustic Surface Elevation Dynamics sensors (O-SED and A-SED) around the edge of natural and semi-natural salt marshes. Sensors were installed at vegetated and unvegetated measurement station Field observations from 22 O-SEDs deployed at 4 well-established natural salt marshes in the Western Scheldt estuary and 4 O-SEDs at a well-established semi-natural salt marsh in the Wadden Sea were reanalyzed. Six novel A-SEDs were deployed at a pioneer semi-natural salt marsh in the Ems-Dollard Estuary. The measurement duration at all salt marshes was at least 1 year. The A-SED sensor was successfully validated against manual measurements. Furthermore, vegetation data and water level data were obtained. No significant difference was observed between natural and semi-natural salt marshes. However, a significant difference between vegetated and unvegetated measurement stations for short-term bed level dynamics was observed. Vegetation was found to be present at locations restricted by short-term bed level variability smaller than or equal to 12 mm, emphasizing the presence of a short-term threshold. Although trends in the non-growing season were significantly different between vegetated and unvegetated stations, seasonal thresholds for vegetation presence were not found. The findings imply that knowledge of bed level-dynamics in well-established natural marshes can be used to predict vegetation presence in constructed semi-natural marshes. The importance of local short-term dynamics for vegetation presence instead of longer-term dynamics highlights possibilities for developing favorable conditions for vegetation presence in marsh restoration projects and the construction of new salt marsh ecosystems.

Intertidal salt marshes provide a range of valuable ecosystem services which typically increase with marsh width. Understanding the drivers for salt marsh expansion versus retreat is thus key to managers. Previous research highlights the influence of short-term (daily/event) bed level dynamics on germination and establishment and subsequent vegetation presence. However, more recent literature suggests the importance of medium-term seasonal bed level dynamics on viable seed availability and subsequent vegetation presence. This study aims to assess event-based and seasonal bed level dynamics for vegetation presence in natural and semi-natural salt marshes and to provide generic thresholds for vegetation presence. To gain insight into bed level dynamics, data was used from autonomous Optical and Acoustic Surface Elevation Dynamics sensors (O-SED and A-SED) around the edge of natural and semi-natural salt marshes. Sensors were installed at vegetated and unvegetated measurement station Field observations from 22 O-SEDs deployed at 4 well-established natural salt marshes in the Western Scheldt estuary and 4 O-SEDs at a well-established semi-natural salt marsh in the Wadden Sea were reanalyzed. Six novel A-SEDs were deployed at a pioneer semi-natural salt marsh in the Ems-Dollard Estuary. The measurement duration at all salt marshes was at least 1 year. The A-SED sensor was successfully validated against manual measurements. Furthermore, vegetation data and water level data were obtained. No significant difference was observed between natural and semi-natural salt marshes. However, a significant difference between vegetated and unvegetated measurement stations for short-term bed level dynamics was observed. Vegetation was found to be present at locations restricted by short-term bed level variability smaller than or equal to 12 mm, emphasizing the presence of a short-term threshold. Although trends in the non-growing season were significantly different between vegetated and unvegetated stations, seasonal thresholds for vegetation presence were not found. The findings imply that knowledge of bed level-dynamics in well-established natural

INTRODUCTION
Salt marshes have proven to deliver a range of valuable ecosystem services, like providing habitats to unique flora and fauna, carbon sequestration and coastal protection by wave attenuation (e.g., Duarte et al., 2013;Möller et al., 2014;Zhu et al., 2020). More recently, the long-term coastal protection services of salt marsh size has been quantified, by analyzing the relations between the width of the salt marsh and its wave attenuating capacity through time Vuik et al., 2019). Such long-term analyses are needed for the successful implementation of salt marshes as nature-based solutions (NbS) in coastal management, as coastal defense schemes are built to last for decades (Borsje et al., 2011;Temmerman et al., 2013). Long-term (years to decades) lateral dynamics of the marsh edge that determine the width of a marsh via marsh expansion or retreat (Harmsworth and Long, 1986;Van der Wal et al., 2008) have been related to various factors. For example, at the large-scale marsh expansion and retreat has been related to changes in sea level (Allen, 2000), changes in wind and waves (Schwimmer, 2001;van der Wal and Pye, 2004), channel deepening and migration (Pringle, 1995;Pye, 1995;Cox et al., 2003), sediment supply (Ladd et al., 2019) and sedimentation and erosion processes (van der Wal et al., 2002). At a local scale the lateral dynamics of the marsh edge have been related to factors such as seed availability (Rand, 2000;van Regteren et al., 2017van Regteren et al., , 2020b short-term sediment dynamics Willemsen et al., 2018) and soil properties (Wolters and Bakker, 2002).
Wave attenuation by marshes, as well as the buildup of the marsh soil, critically depends on the density of the vegetation (e.g., Barbier et al., 2008;Bouma et al., 2014). The establishment of pioneer plants is the first stage of making the environment more suitable for subsequent plant colonization and marsh development by locally decreasing hydrodynamic energy (van Regteren et al., 2020a). However, vegetation can only colonize locations where, amongst other factors, hydrodynamic forces and morphodynamic stresses remain below certain critical thresholds Bouma et al., 2016). That is, given fulfillment of local biological conditions such as seed availability, an inundation-free period is required for initial establishment, followed by a calm period with rather low waves and limited bed level dynamics for subsequent growth (Hu et al., 2015b;Cao et al., 2018;Poppema et al., 2019), which is described as a Window of Opportunity (WoO; Balke et al., 2014). These WoOs capture the first weeks/months of vegetation establishment and growth (Cao et al., 2019;Poppema et al., 2019). The WoO theory is primarily focused on the growing season during which the seedlings establish . In addition to the growing season, year-round (including the non-growing season) bed level change has been observed to be an important factor in driving the lateral dynamics of the salt marsh edge van Regteren et al., 2019). Extending on the WoO theory, these studies suggest that year-round disturbances are also important for the presence of vegetation.
Semi-natural salt marshes are marshes that are (partly) created by humans for their ecosystem services. They often consist of intertidal areas that are made suitable for vegetation establishment by engineering measures that reduce local disturbances. However, initial conditions differ, depending on the design and construction methods applied. For example, semi-natural marshes can be obtained by artificially elevating the intertidal area with dredged sediments (Ford et al., 1999;Schrift et al., 2008) and/or by reducing hydrodynamic and morphodynamic stresses by adding wave-attenuating structures . Although the number of semi-natural (constructed) salt marshes is increasing (Maris et al., 2007;Mazik et al., 2007;Temmerman et al., 2013;Rahman et al., 2019), biophysical data on constructed salt marshes remain scarce. While daily to seasonal morphodynamics have been observed to drive the lateral location of the marsh edge in natural salt marshes (e.g., Bouma et al., 2016;Willemsen et al., 2018), observations of similar dynamics in constructed salt marshes are lacking. To be able to efficiently create constructed salt marshes for their ecosystem services, it is key to quantify morphodynamics and vegetation presence, and to compare those dynamics with those observed in natural salt marshes. We hypothesize that event-driven (short-term) and seasonal (medium-term) bed level dynamics do not differ between natural and semi-natural constructed salt marshes.
Continuous monitoring of bed level dynamics is required to enable the quantification of morphodynamics at a temporal resolution of hours, associated with the impacts of storm events. Simultaneously, these morphodynamics need to be monitored for a duration of multiple seasons to years to uncover seasonal bed level dynamics. Periodic bed level changes have been monitored extensively, by manual monitoring techniques using erosion pins (Stokes et al., 2010), steel reference plates (Andersen et al., 2006), sediment traps (Moskalski and Sommerfield, 2012), accretion poles (O'Brien et al., 2000) and sedimentation-erosion bars . However, the use of autonomous systems (such as the Photo-Electronic Erosion Pin) to monitor bed level dynamics at a high temporal resolution (hours) over periods of seasons to years remains scarce due to instrument limitations and costs (Lawler, 2008;Hu et al., 2015a). More recently, the Optical Surface Elevation Dynamics sensor (O-SED) was developed and applied for continuous monitoring of bed level dynamics in the intertidal area, when the sensor is not submerged, at a temporal resolution of hours and for a duration of multiple months (Hu et al., 2017(Hu et al., , 2020Willemsen et al., 2018). This instrument was further developed into an Acoustic Surface Elevation Dynamics (A-SED) sensor, now able to capture bed level change when submerged.
In this study, we used data of both O-SED (26) and novel A-SED (6) sensors around the edge of natural and semi-natural salt marshes in order (1) to assess the magnitude of eventbased and seasonal bed level dynamics in natural and seminatural (constructed) salt marshes and (2) to identify generic thresholds of event-based and seasonal bed level dynamics for vegetation presence, both in natural and semi-natural salt marshes. These thresholds quantify maximum accretion and/or erosion vegetation around the salt marsh edge can withstand. The structure of this paper is as follows. In section "Materials and Methods" the study sites are introduced, followed by a description of the state-of-the-art instruments deployed for yearround field data collection of high-resolution bed level dynamics. Section "Results" presents the results of data analyses based on the collected field data and quantifies the bed level change limits for vegetation growth across the study sites. In section "Discussion" the main findings of this paper are discussed, followed by the main conclusions in section "Conclusion."

Study Sites
The study sites are located around the North Sea coast of the Netherlands at the Western Scheldt estuary, Wadden Sea and Ems-Dollard estuary (Figure 1). Field data were previously collected at four natural salt marshes in the Western Scheldt, that have been thriving for decades and at one artificial seminatural salt marsh in the Wadden Sea with brushwood structures also existing for decades . This data was reanalyzed. State-of-the-art A-SED sensors were deployed at a second semi-natural salt marsh in the Ems-Dollard Estuary that has been constructed in 2018 by creating an elevated bed protected by brushwood structures.

Natural Salt Marshes
The four natural salt marsh sites are located in the Western Scheldt estuary, which is located in the southwestern delta of the Netherlands (51 • 20 N; 4 • E; Figure 1 and Table 1). Two sites, Zuidgors Hellegat and Zimmerman, are located at the northern shores of the Western Scheldt, approximately 20 and 50 km from the mouth of the estuary, respectively. The other two sites in the Western Scheldt, Paulina and Hellegat, are located at the southern shores, approximately 15 and 27 km from the mouth, respectively. The sites at the northern shore are exposed, whereas the sites at the southern shore are more sheltered from the prevailing south-westerly winds (Callaghan et al., 2010). The sites Zuidgors and Paulina, are saline at 20-25 psu, whereas the two study sites further inland, Hellegat and Zimmerman, are brackish at 15-20 psu (Damme et al., 2005). The mesotidal to macrotidal Western Scheldt estuary is tide dominated with a semi-diurnal tide (Baeyens et al., 1997). The spring tidal range at Paulina, the study site closest to the mouth of the Western Scheldt estuary, is approximately 4.4 m and increases to 5.5 m at Zimmerman, the study site closest to the landward boundary of the estuary ( Van der Wal et al., 2008). The grain size of the top layer of the bed (D 50 ) varied between 18 and 132 µm ; Table 1). In the more saline marshes in the west of the estuary, pioneer species such as perennial common cord grass (Spartina anglica) and the annual glasswort (Salicornia spp.) are occurring. Well-established vegetation, such as common saltmarsh grass (Puccinellia maritima), annual seablite (Suaeda maritima), and sea aster (Aster tripolium) are occurring higher in those marshes. At the most eastern, more brackish study sites, well-established seaside bulrush (Bolboschoenus maritimus) and common reed (Phragmites australis) are occurring as well as common cordgrass (Spartina spp.) ( Van der Wal et al., 2008).

Semi-Natural Salt Marshes
The two semi natural salt marshes are located in the Wadden Sea and Ems-Dollard estuary. The semi-natural salt marsh site Uithuizen is located in the Wadden Sea area, close to the mouth of the Ems-Dollard estuary (53 • 27 N, 6 • 39 E; Figure 1 and Table 1). The Wadden Sea is a mesotidal to macrotidal estuary (Dieckmann et al., 1987), with a tidal range of approximately 3.0 m at Uithuizen. The salinity near Uithuizen is approximately 30 psu (Zimmerman, 1976;Ysebaert et al., 2016). The D 50 at the study site was approximately 87 µm (Folmer et al., 2017; Table 1). Salt marsh development in this particular marsh is supported by brushwood structures (van Loon-Steensma, 2015). At this site, common glasswort (Salicornia europaea), common salt marsh grass (Puccinellia maritima) and common cordgrass (Spartina anglica), are pioneer species occurring in the lower salt marsh. Vegetation species occupying the higher marsh are couch grass (Elymus athericus), sea aster (Aster tripolium), and seepweed (Suaeda maritima).
The semi-natural Marconi salt marsh (53 • 19 39.33 N, 6 • 56 40.75 E; Figure 1 and Table 1) is located in the Ems-Dollard estuary, at Delfzijl. The mesotidal Ems-Dollard estuary is directly connected with the Wadden Sea estuary. The tidal range near Delfzijl is approximately 3.1 m, with a spring tidal range of 3.4 m (Dillingh, 2013). The salinity in the Ems-Dollard increases from southeast to northwest and is generally below 20 psu (Esselink et al., 1998), with approximately 23 psu locally near the Marconi salt marsh. The Marconi salt marsh was developed in 2018 at the seaward side of the jetty of the Port of Delfzijl. Mud (silt and clay) was mixed through the top layer of a sand fill on the foreshore of the jetty. The mud was collected during a local port extension project and temporarily stored in a land depot. The percentage of mud in the sand/mud mixture was varied intentionally between different sections of the salt marsh (ranging between 5 and 50%; Figure 1). The D 50 at the top layer of the sections were measurement stations were located varied between 111 and 213 µm (De Vries et al., 2021; Table 1). To protect establishing vegetation from excessive hydrodynamic energy, permeable brushwood dams FIGURE 1 | Study sites around the North Sea area (A) indicated with a red dot (B). Locations where water levels were continuously monitored are indicated with a blue dot and the location were wind data were collected is indicated with a black dot (B). The measurement stations (numbered) for bed level dynamics are marked with green (vegetated) and white (unvegetated) dots for each of the study sites (C-H). Source of background maps (C-G) is Google Maps, source of background map (H) is MapTiler and OpenStreetMap contributors.
were constructed surrounding the different salt marsh sections. Moreover, to enhance vegetation growth, Salicornia procumbens was sown in part of the salt marsh sections. After its first year, vegetation species established at the Marconi salt marsh were mainly the pioneer and lower marsh species glasswort (Salicornia spp.), seaside bulrush (Bolboschoenus maritimus), spear-leaved

Vegetation Data
At the natural salt marshes in the Western Scheldt and semi natural salt marsh in the Wadden Sea, where vegetation has already been developing for decades, vegetation presence was determined for each of the measurement stations. Measurement stations where vegetation was absent during the measurement periods were defined as bare tidal flat, whereas stations where vegetation was present were defined as vegetated salt marsh. Vegetation cover at the salt marsh stations was characterized as pioneer, low marsh and high marsh, each with a vegetation cover of at least 5%. A similar approach was applied at the constructed Marconi salt marsh. Vegetation was not present when the measurement stations were installed, due to recent construction of the seminatural salt marsh. Measurement stations at Marconi were setup around the local mean high water level (1.4 m +NAP). Vegetation cover was monitored at the end of the growing season (September) to be able to observe annual species as well, and vegetation density was estimated using a decimal scale (Londo, 1976) in permanent quadrats (PQ's) of 2 m × 2 m. At the end of the growing season, measurement stations were characterized either by vegetation presence or absence.

Water Level and Wind Data
Water level data were obtained from Rijkswaterstaat (Dutch waterways and public works Authority), for consistency across all study sites. Water level data were obtained at a 10 min interval at measurement stations near the study sites: Terneuzen, Bath, Hansweert, Uithuizen and Delfzijl (Figure 1). The surface elevation of the measurement stations, relative to NAP (Dutch Ordnance Datum), was initially measured with an RTK-DGPS for determining inundation periods of the individual measurement stations. The RTK-DGPS has a vertical accuracy of approximately 2 cm. Hourly wind data (wind speed and wind direction) were obtained from KNMI (The Royal Netherlands Meteorological Institute) measurement stations near the natural salt marsh Zuidgors and the constructed salt marsh Marconi, Vlissingen and Nieuw-Beerta, respectively.

Bed Level Dynamics
Bed level dynamics data were collected or reanalyzed at all study sites, both at locations with and without vegetation presence (Figure 1). All measurement stations were located close to the vegetation edge (or predicted vegetation edge at the end of the growing season), to assess the isolated influence of bed level dynamics on the occurrence of salt marsh vegetation, as well as the similarity of bed level dynamics at natural and semi-natural salt marshes. Field observations from a total of 32 measurement stations across the six study sites were analyzed, each for at least a 1 year duration ( Table 1).
Bed level changes at the natural salt marshes in the Western Scheldt estuary were collected previously using the Optical Surface Elevation Dynamics (O-SED) sensor . This stand-alone sensor captures bed level change by using a measurement array of light sensitive cells (i.e., phototransistors) and has been successfully applied to quantify salt marsh dynamics in previous studies (Hu et al., 2015a(Hu et al., , 2017. The sensors were installed firmly in the bed, with half of the measurement array of 400 mm below the surface and half of the array extending above the sediment surface. The instruments were serviced, cleaned and data were downloaded every 2-6 months The sensors were pushed in or pulled out of the bottom after excessive sedimentation and erosion. The observations were corrected for this action during postprocessing. A single surface elevation measurement was collected every 30 min. Raw data were converted to bed level change data using an existing post-processing script . The post-processed data have previously been validated successfully against sedimentation-erosion bars (Hu et al., 2015a) and erosion pins , with an accuracy of the obtained bed level changes of approximately 5.0 mm (Hu et al., 2015a). Data was reanalyzed specifically for the current study.
Bed level changes at the semi-natural salt marsh Uithuizen at the Wadden Sea coast were also previously collected using the O-SED sensor . Data collection, post-processing and reanalysis, was equal to collection, postprocessing and reanalysis for marshes in the Western Scheldt. At the semi-natural constructed Marconi salt marsh, bed level change data were collected using state-of-the-art Acoustic Surface Elevation Dynamics (A-SED) instruments (Figure 2). This recently developed instrument measures the distance between the bed and the instrument through the echo of an acoustic signal (Appendix 1). The sensor operates stand-alone without requiring external power and data-logging systems. Furthermore the instrument is relatively inexpensive (±1,000 euro) compared to other (acoustic) measurement instruments (see Hu et al., 2015a). The instrument is set-up to be able to collect measurements when the sensor of the instrument is fully inundated. A burst of eight samples (sampling frequency of 5 Hz) was stored at a measurement interval of 5-15 min. The instruments were installed downward looking, with the sensor head aimed vertically at the bottom (Figure 2). The instruments were serviced, cleaned and data were downloaded every 2-3 months. A post-processing script was developed in Python to convert the acoustic return signals to bed level change data (Appendix 1). The measurements were validated with manual readings of the height above bed (HAB) of the sensors every time the instruments were serviced.

Data Analysis
The distribution of the magnitude of bed level change per measurement (with a measurement every 5-30 min) was determined per study site both for the stations were vegetation was present and absent, by calculating the difference of subsequent measurements. The standard deviation of every full time series of bed level dynamics was calculated to capture shortterm event-based dynamics, since larger bed level change was observed to coincide with larger waves and stronger winds (c.f. Hu et al., 2015a;Willemsen et al., 2018). Medium-term seasonal bed level change was captured by calculating the linear leastsquares regression of the observed bed level dynamics per season, with the slope of the regression representing the linear rate of change of the bed level over a single season. For this analysis, the seasons were defined as the growing season (April-September) and non-growing season (October-March). If gaps in the SED data appeared for three or more consecutive months, either in the growing or in the non-growing season, the seasonal rate of change of the station was discarded. Additionally, local water level data were compared with RTK-DGPS-derived bed level data of the measurement stations, to compute the time each site was submerged/emerged.

Validation of the Accuracy of the Novel Sensors and the Processing Software
Bed level change measurements collected with the A-SED sensors at all measurement stations at Marconi and the post-processing script (Appendix 1) were validated with manual measurements, analog to what was done earlier for the O-SED (Hu et al., 2015a;Willemsen et al., 2018). An average height above bed (HAB) was calculated for each burst of eight individual A-SED measurements. The raw data each of those eight measurements were first converted to a height above bed (HAB), which were subsequently averaged. The HAB of the A-SED was also manually measured upon installation and retrieval of the sensors, every Frontiers in Marine Science | www.frontiersin.org 2-3 months. These manual measurements were compared with the first instrument measurement after installation and the last measurement before retrieval, respectively (Figure 3).
A direct comparison between the A-SED data and the manually obtained reference measurements provides a nearunity correlation: the least-squares regression line of the A-SED data and the manually obtained measurements has a slope of 1.0001 with an R 2 of 0.99 (Figure 3). The height measured by the A-SED sensors was systematically higher than the manual observations by 2.13 cm. This offset was subsequently applied as a fixed calibration constant in the post-processing of the A-SED data. For the current analysis, however, this constant offset is not affecting the results, since all A-SED observations are converted into bed level change relative to the initial measurement of the time series. The accuracy of the sensor is defined as the absolute error between the A-SED sensor measurement and the true value assumed to be measured manually (c.f. Hu et al., 2015a). Based on all manual measurements, the accuracy is 7.6 mm with a standard deviation of 6.1 mm. The accuracy of the O-SED sensor was observed to be 5.0 mm (Hu et al., 2015a), comparable to the A-SED sensor.

Vegetation Presence
Vegetation was present, for the full bed level dynamics measurement period, at 12 of the 22 measurement stations located at the natural salt marshes in the Western Scheldt (Zuidgors, Paulina, Hellegat, and Zimmerman). The vegetation cover at the vegetated stations was at least 5%. At the semi-natural salt marsh in the Wadden Sea vegetation was present at 3 of the 4 measurement stations (cover >5%) (Figure 1). Vegetation at the semi-natural salt marsh Marconi was still developing, due to its recent construction. Vegetation at the end of the growing season in September 2020 was present at 3 of the 6 measurement stations and consisted mainly of the annual Salicornia spp. The vegetation density at measurement stations 1, 2, 3, 4, 5, and 6 was 0, 0, 0, 10, 1, and 20%, respectively (Figure 1).

Inundation Time of the Measurement Stations
All well-established natural salt marshes in the Western Scheldt generally showed an inclining elevation from the most seaward measurement station toward the most landward measurement station. This is reflected in reducing inundation times at these measurement stations (Figure 4). The lowest elevation/inundation level where vegetation is present was strongly site-specific (Figure 4; dashed lines). In the Western Scheldt estuary, where four natural salt marshes were located, vegetation extended more seaward to lower elevations and was able to withstand shorter dry periods (i.e., longer inundation) at the southern shores (Paulina and Hellegat), compared to the northern shores (Zuidgors and Zimmerman). This might be explained by the prevailing wind conditions, adding local disturbances: Paulina and Hellegat are sheltered from the prevailing wind conditions, whereas Zuidgors and Zimmerman are more exposed to the prevailing southwesterly winds.
At the semi-natural salt marsh Uithuizen in the Wadden Sea the inundation period increased from the seaward to landward measurement stations, similar to the observations in the Western Scheldt estuary. At the semi-natural (recently constructed) Marconi salt marsh in the Ems-Dollard estuary, the inundation duration is longer at the measurement stations in front of the brushwood dams (Figure 4; Marconi stations 1 and 2). In general, the dry period at the measurement stations sheltered by the brushwood dams is longer, enabling vegetation to establish (Figure 4; stations 3, 4, 5, and 6). Nevertheless, even though the elevation of the most eastward measurement station sheltered by the brushwood dams (station 3) exceeds the local lowest elevation for vegetation to grow, vegetation was still not able to grow there. This might be explained by both the location of the test section being near to the edge of the constructed salt marsh, thereby having slightly higher hydrodynamic energy and/or by the low clay content of this test section (5%), resulting in less favorable growing conditions.
Although similar patterns were observed at all study sites, the inundation threshold for vegetation to be present was different. So the results show that inundation time can explain the presence of vegetation locally, within each study site. However, this physical parameter, strongly related to elevation, does not provide a generic threshold to the presence of vegetation for all study sites around the North Sea.

Bed Level Change
Between the time series of bed level dynamics (time scale of minutes), both deposition (positive) and erosion (negative), at all study sites there was a significant difference [twoway ANOVA; F(5,552643) = 7088, p < 0.001], indicating unique dynamics for all individual sites. Between the vegetated and unvegetated measurement stations, over all study sites there was a significant difference in the full bed level dynamics time series [F(1,552643) = 709, p < 0.001], highlighting different dynamics at measurement stations with and without vegetation.
The frequency distribution of absolute bed level change measurements (magnitude; both accretion and erosion are positive), both for stations with and without vegetation, showed some similarities (Figure 5). Small bed level changes were observed most often at all sites. The frequency of bed level change declined with increasing magnitudes of bed level change, whereafter a trend break was observed (at bed level change between 5 and 10 mm) toward a steeper decline (Figure 5). The shape of the frequency distribution before the trend break varied largely between the different study sites as well as the vegetated and unvegetated stations. Largest differences between the vegetated and unvegetated distributions were observed at the natural salt marsh Zimmerman and seminatural marshes Uithuizen and Marconi, although the frequency of larger bed level changes at unvegetated stations exceeded those at the vegetated stations at Zimmerman and Uithuizen contrary to Marconi.
The mean and 90th percentile of absolute bed level change (magnitude; both accretion and erosion are positive) at the stations without vegetation did not exceed values at the stations with vegetation at all sites, contrary to expectations. The mean and 90th percentile at the unvegetated stations at Paulina and Marconi did not exceed the values at the vegetated stations at these sites, whereas the opposite was observed at the other study sites. For example, the mean bed level change for Hellegat over the full time series was 0.9 mm at the unvegetated and 0.7 mm at the vegetated stations, whereas for Marconi the mean bed level change was 1.1 mm at the unvegetated and 1.6 mm at the vegetated stations. No generic mean or 90th percentile bed level change threshold was found that distinguished between vegetated and unvegetated for all study sites.

Event-Based Bed Level Dynamics
Measured bed level change, both deposition (positive) and erosion (negative), was converted to a standard deviation to assess the magnitude of bed level change due to, e.g., windwave events. As an example bed level change data at the natural site Zuidgors were compared with locally occurring south and south-westerly wind conditions (associated with the longest fetch at this site), to highlight differences in bed level dynamics during storm and calm conditions that were captured with the (A-SED) sensors. The standard deviation of bed level change during a period with storm conditions (e.g. onshore wind speeds exceeding 75 km/h) generally was larger than the standard deviation during calm conditions, both at the unvegetated and vegetated measurement stations (Figure 6). Similar variability in the bed level dynamics due to storm conditions was observed at the semi-natural site Marconi. Storm conditions caused by locally occurring northerly onshore wind conditions (associated with the longest fetch for this site) were not present in the data. Nevertheless, bed level dynamics during onshore windy periods (periods with largest wind speeds, exceeding 40 km/h) were substantially larger than during calm periods (Figure 7), assuming no bed level change when stations were not inundated (Figure 7). At all study sites, unvegetated measurement stations showed substantially larger dynamics than measurement stations where vegetation was present (Figures 6-8). The standard deviation of bed level dynamics for the full measurement period generally showed larger values at the unvegetated stations than the vegetated stations across all measurement stations at the six study sites (Figure 8). A significant difference was observed between the measurement stations where vegetation was present and absent [two-way ANOVA; F(1,29) = 45, p < 0.001]. A significant difference for short-term event-based bed level change between measurement stations at natural and seminatural salt marshes was not found. The standard deviations of bed level dynamics for the full measurement period at all vegetated stations at the natural and semi-natural salt marshes were not exceeding 12 mm (Figure 8), indicating a threshold for vegetation presence. Event-based bed  level change at unvegetated measurement stations was generally larger, except at one measurement station at Marconi (Figure 8). Supported by the significant difference between event-based bed level change at vegetated and unvegetated measurement stations: for vegetation to be present the maximum short-term bed level change should not exceed 12 mm.

Seasonal Trends in Bed Level Change
To extend from short-term, event-based bed level change toward medium-term seasonal trends in bed level change, seasonal rates of bed level change were computed (Figures 9, 10). In general, vegetated areas were accreting (up to 111 mm/season) or slightly eroding (up to 30 mm/season) during the growing season. Surprisingly, rates of change at the unvegetated measurement stations were larger during the growing season, both for accretion (up to 155 mm/season) and erosion (up to 90 mm/season) (Figure 10). During the non-growing season, the rates of change at the vegetated stations were of similar size as during the growing season (up to 52 mm/season accretion and up to 33 mm/season erosion). Dynamics at the unvegetated measurement station were generally lower in the non-growing season compared to the growing season (up to 102 mm/season) ( Figure 10). Although the rates of change at the vegetated stations seemed generally lower than at the unvegetated stations at each individual study site (Figure 10), a significant difference between all vegetated and all unvegetated measurement stations was only observed during the non-growing season [two-way ANOVA; F(1,23) = 22, p < 0.001]. No significant difference was observed between vegetated and unvegetated measurement stations during the growing season. Dynamics at vegetated and unvegetated stations both in the growing and non-growing season overlapped such that a general threshold for vegetation presence was not found.
Rates of change at the natural marshes ranged from 90 mm/season erosion to 155 mm/season accretion in the growing season and from 102 mm/season erosion to 25 mm/season accretion in the non-growing season. Observed differences at the semi-natural marshes ranged from 80 mm/season erosion to 70 mm/season accretion in the growing season and from −38 mm/year sedimentation to 52 mm/season accretion during the non-growing season, indicating a larger range over the growing season. No significant difference was observed between natural and semi-natural salt marshes both at the medium-term growing and non-growing season.

Combining Short-Term Event-Based Bed Level Dynamics and Medium-Term Seasonal Trends in Bed Level Change
When plotting event-based and seasonal bed level dynamics against each other, measurement stations where vegetation is present are all grouped in a parameter space delimited by relatively low event-based bed level change (i.e., standard deviation of bed level dynamics time series (Figure 11). We did not find specific seasonal limitations preventing vegetation presence. Vegetation was present at a short-term event-based bed level variability smaller than or equal to 12 mm, exceedance of this threshold was found to coincide with the absence of vegetation. This threshold was found to be valid both for natural and semi-natural salt marshes.

DISCUSSION
The current study highlights that (1) physical observations of bed level change at different timescales are similar at natural and semi-natural salt marshes including the recently constructed semi-natural Marconi salt marsh. Short-term event-based bed level change (2) was observed to be significantly different at vegetated and unvegetated measurement stations, and (3) limited the presence of vegetation. Although seasonal trends were significantly different between vegetated and unvegetated measurement stations during the non-growing season, seasonal thresholds for vegetation presence were not observed. Hence, the threshold for event-based bed level dynamics was sufficient for explaining vegetation presence.

Monitoring High-Resolution Bed Level Dynamics
A unique dataset of high-resolution bed level dynamics was compiled using state-of-the-art A-SED and O-SED measurement instruments. Bed level dynamics were analyzed at 32 measurement stations distributed across six study sites at the North Sea coast of the Netherlands. Previous studies have highlighted the importance of field observations of bed level dynamics with a high vertical spatial and temporal resolution, for example quantifying the influence of storms and vegetation establishment (Hu et al., 2015a(Hu et al., ,b, 2020Willemsen et al., 2018). The current study extends on these previous observations by introducing and validating the A-SED sensor. Unlike manual observations and measurements with other sensors such as the O-SED (e.g., Lawler, 1991;O'Brien et al., 2000;Stokes et al., 2010;Baptist et al., 2019), the A-SED sensors allow for autonomous and continuous measurement of bed level changes when measurement stations are submerged. This allows the new A-SED sensors to monitor bed level change in real-time under hydrodynamic exposure, as opposed to these dynamics being derived from comparing pre-and post-hydrodynamic exposure measurements of the bed level during low-tide manually or with autonomous instruments such as the O-SED. The A-SED sensor is not restricted by light conditions, which is an advantage over previous optical sensors (Lawler, 1991;Hu et al., 2015a). Accuracy of the A-SED (7.6 mm) and O-SED (5.0 mm) was rather similar (Hu et al., 2015a). The A-SED measurements and manual measurements were not taken at exactly the same time, since the A-SED measures when submerged (requiring a water depth of approximately 0.3 m for the measurement head to be submerged) and the manual measurements were obtained when the stations were emerged, which might have caused the small difference between both measurements, thereby decreasing the accuracy. Nevertheless, the rather good agreement between both measurements suggest that the A-SED sensor, similar to the O-SED sensor, is a reliable instrument for continuous and autonomous measurements of bed level change.

Morphodynamic Disturbances for Vegetation Presence
In general, event-based bed level dynamics and seasonal rates of change were greater at the unvegetated locations at each site individually. The stabilizing effect of vegetation on bed level dynamics (Gacia and Duarte, 2001;Möller et al., 2014), was present in the well-established marshes in the Western Scheldt and Wadden Sea, since established annual and perennial vegetation species were present at the start of the measurements. However, vegetation was not yet present at the pioneer seminatural Marconi salt marsh at the start of the measurements. So at this site, there was only an effect of the bed level dynamics on the presence of vegetation whereas there was no or negligible feedback of the vegetation on bed level dynamics. Despite the different feedback mechanisms in well-established versus   pioneer marshes, both event-based and seasonal-based bed level dynamics were similar at all study sites. Moreover, vegetated and unvegetated stations at all salt marshes were characterized by a single event-based threshold for vegetation presence. At Marconi a single measurement station where vegetation presence could be expected by its short-term dynamics, was unvegetated (Figure 11). This might be explained by the fact that Marconi was a recently constructed marsh and the marsh profile still is developing toward a dynamic equilibrium, or if exposure is increasing, a persistent erosive trend will prevent for the presence of vegetation (Figure 11). This development and/or trend might be represented in the rate of change during the non-growing season. Still, bed level dynamics along the elevational gradient at the edge zone of well-established (natural) marshes does give insights in maximum dynamics for vegetation presence at semi-natural salt marsh construction and restoration sites.
Previous studies suggest that colonization of the bare tidal flat by pioneer vegetation depends on short-term morphodynamic dynamics (Balke et al., 2011;Hu et al., 2015b;Silinski et al., 2016;Poppema et al., 2019). Short-term bed level dynamics can prevent successful germination and dislodge young seedlings if roots are not anchored sufficiently (Hu et al., 2015b). The current study combines the studies on short-term bed level dynamics with studies suggesting medium-term seasonal and annual bed level change as explanation for vegetation presence van Regteren et al., 2019). For example erosion of the bed in the winter season might remove viable seeds of annual species, thereby limiting vegetation presence in the subsequent growing season . Interestingly, a prohibitive effect of accretion on the presence of vegetation was not observed at our study sites. Our findings show that vegetation presence is mainly driven by short-term eventbased bed level dynamics. This might indicate that vegetation presence was not limited by viable seed availability at our study sites. Only a single unvegetated measurement station at the Marconi marsh might be hampered by a medium-term trend at the non-growing season.

Implications for Salt Marsh Construction and Restoration
While the salt marsh edge at natural salt marshes is relatively stable over the period of a year (Allen, 2000;Van der Wal et al., 2008;Singh Chauhan, 2009;Willemsen et al., 2020), the seminatural constructed Marconi salt marsh was still developing, i.e., vegetation had first appeared only 1 year before starting the measurements. Similarities in variability of bed level dynamics around the marsh edge, both at the natural and semi-natural salt marshes, imply that bed level dynamics around natural salt marshes can be used as a proxy for the outline of salt marsh construction and restoration schemes. Locations where more hydrodynamic energy is present, leading to increasing bed level dynamics, require a higher elevation for vegetation to develop, or need additional protection (Cao et al., 2018(Cao et al., , 2019Willemsen et al., 2018). Similar bed level dynamics at the salt marsh edge suggest that disturbance-decreasing measures at the constructed Marconi salt marsh have been successful in creating sheltered areas to enhance vegetation presence (c.f. Fivash et al., 2021;Gijón Mancheño et al., 2021).
General parameterizations, such as the morphodynamic thresholds for vegetation presence presented in this study, can provide key insights for predictions of vegetation growth in intertidal areas. The current study highlights that local shortterm bed level dynamics, that may be altered by local (temporary) management measures, control vegetation presence (c.f. Hu et al., 2021). Balke et al. (2016) introduced a generic hydrodynamic threshold for the presence of vegetation at the salt marsh edge, based on tidal inundation frequency. Combined, these thresholds can be used (1) to predict salt marsh development, and (2) to design nature-based solutions incorporating salt marshes including the requirement of sheltering structures to prevent structural medium-term erosion. Moreover, the use of thresholds instead of process-based modules for vegetation development in biogeomorphological models (Kirwan and Murray, 2007;Temmerman et al., 2007;Tonelli et al., 2010;Hu et al., 2015b;Best et al., 2018;Schwarz et al., 2018;Poppema et al., 2019;Willemsen et al., 2022), provides a simple rule-based approach for the simulation of vegetation development in numerical models (Huang et al., 2008), decreasing their complexity without assuming static vegetation (Horstman et al., 2015;Willemsen et al., 2016). Hence, existing numerical hydro-morphodynamic models for designing nature-based solutions (e.g., Siemes et al., 2020), can include the development of salt marsh vegetation by only assessing whether intertidal environments are suitable for vegetation growth at different timescales, without increasing model complexity and computational cost. Obviously, a second step is to incorporate the feedbacks between the vegetation present and the hydro-and morphodynamics in its surroundings.
When aiming at salt marsh conservation, restoration or construction, e.g., for optimizing ecosystem services such as coastal protection (Möller et al., 2014) and carbon sequestration (Duarte et al., 2013), it is key to monitor the state of the ecosystem. In order to preserve the size of the ecosystem or to allow for its development, bed level dynamics induced by the combined action of tides and waves and modulated by the presence of vegetation should not exceed the eventbased threshold that was identified to prevent vegetation from establishment and growth. The quantification of thresholds for bed level dynamics, combined with actively monitoring existing salt marshes and salt marsh restoration projects, provides critical insights on whether salt marsh systems are stable or will be accreting or eroding. Future research on longer-term periods than observed in the current study (years to decade) are essential for assessing the influence of climate change on long-term bed level change potentially hampering vegetation growth. Based on event-based and longer-term thresholds, adaptive management strategies can be developed. Once bed level dynamics come close to thresholds for vegetation presence, additional measures, such as construction of structures reducing physical stress and sediment nourishments, can be applied to protect the ecosystem and its services.

CONCLUSION
Present results suggest that event-based bed level dynamics can explain vegetation presence in salt marshes distributed across the Western Scheldt estuary, Wadden Sea and Ems-Dollard Estuary along the Dutch North Sea coast. Short-term bed level change is observed to provide a critical threshold for the presence of salt marsh vegetation. Vegetation at the well-established salt marsh was found to be present at event-based bed level variability smaller than or equal to 12 mm. The importance of local short-term dynamics for vegetation presence instead of longerterm dynamics highlights possibilities for developing preferential conditions for vegetation presence. So the limiting role of bed level dynamics on the presence of salt marsh vegetation is key to design and to predict the establishment of both restored and constructed salt marshes. Knowledge of these thresholds is critical for the development of adaptive management strategies for restoration and construction of salt marshes.

DATA AVAILABILITY STATEMENT
Data and scripts in support of this manuscript for the Western Scheldt and Wadden Sea study site and SED data conversion are available via DOI: 10.4121/uuid:c0be318c-9858-4a05-a546-782e29b4abef. Data and scripts for the Ems Dollard study site and ASED data conversion are available via DOI: 10.4121/16782613.

AUTHOR CONTRIBUTIONS
PW and MP conducted fieldwork for the study. MB was fieldwork leader of the constructed salt marsh. TB initiated the development of the optical-SED (O-SED) and acoustic-SED (A-SED) sensors used in this research. PW developed the postprocessing script for the sensors and wrote the article with input from EH, TB, MB, MP, and BB. PW, EH, and BB designed the outline of the manuscript. All authors reviewed, revised, and approved the manuscript.

FUNDING
The Marconi field work was supported by the municipality of Delfzijl with funding from the Dutch Waddenfonds under grant number WF223001. Cash and in-kind co-funding was received from the consortium partners of EcoShape. Contributions by EH and PW were supported by the Mangrove-RESCUE project (Grant No. 15899), funded by the Dutch Research Council (NWO).