Salt marsh establishment in poorly consolidated muddy systems: effects of surface drainage, elevation, and plant age

Conservation and restoration of salt marsh ecosystems are becoming increasingly important because of the many ecosystem services they provide. However, the processes controlling salt marsh establishment and persistence, especially on bare tidal flats in muddy areas, remain unclear. As muddy sediments typically experience a restriction of soil drainage, we expect that a surface drainage relief due to a heterogeneity topography, as might occur on the edge of tidal channels, can facilitate the establishment of salt marsh vegetation on muddy tidal flats. By means of a manipulative field experiment, using “MegaMarsh Organ” mesocosms, we investigated the impact of surface drainage and elevation relative to mean sea level on (1) the survival of Spartina anglica seedlings from three different age classes: 1-yr, 3-month, and 1-week; and (2) the growth performance of mature S. anglica marsh tussocks. S. anglica seedling survival, especially in the establishment phase, was positively affected by better surface drainage, increases of seedling age, and higher elevation relative to mean sea level. That is, the survival rate of S. anglica seedlings at the end of 6th week increased from 0% (at surface water undrained, 1-week, 0 cm elevation) to 94.44% (at surface water drained, 1-yr, 90 cm elevation). In contrast, surface drainage did not affect the performance of large S. anglica marsh tussocks, as only increased elevation relative to mean sea level was shown to affect S. anglica tussock growth in terms of plant height, shoot numbers, and dry biomass. Based on our findings, we proposed a conceptual model to understand how surface drainage-driven feedbacks in a heterogeneous topography may be reinforced to induce salt marsh establishment in muddy systems. Further testing of present hypothesized model would be beneficial for insights into salt marsh establishment


INTRODUCTION
Conservation and restoration of salt marsh ecosystems are becoming increasingly appreciated because of the many vital ecosystem services they provide (Gedan et al. 2009, Barbier et al. 2011, Shepard et al. 2011, Kirwan and Guntenspergen 2012, Kirwan and Mudd 2012, Burden et al. 2013, Schepers et al. 2017). Yet, the decline and degradation of salt marsh ecosystems continue globally, due to the combined threats of sea-level rise and anthropogenic impacts (Silliman et al. 2012, Kirwan and Megonigal 2013, Schepers et al. 2017. Until recently, we still do not sufficiently understand the processes controlling marsh establishment and persistence (Fresiss et al. 2012, Balke et al. 2014, 2016, 2016. Especially for poorly consolidated muddy systems, such lack of knowledge has largely hampered a majority of salt marsh restoration schemes in many locations at different scales (Broome and Herman 2000, Moreno-Mateos et al. 2012, Mossman et al. 2012. Salt marshes are complex bio-geomorphic ecosystems occurring in the dynamic intertidal zone, where they are subject to changing environmental conditions (Adam 2002). Salt marshes have been generally suggested to present habitat transition from low-elevation, non-vegetated tidal flats to high-elevation, vegetated marshes (Marani et al. 2010, Wang and Temmerman 2013, van Belzen et al. 2017). The transition from bare tidal flats to salt marshes is either initiated by seedling establishment or by clonal expansion, possibly after translocation of clonal fragments. The vegetated marsh is stabilized by biogeomorphic positive feedback loops between vegetation growth, sediment trapping, and sediment stabilization (van Wesenbeeck et al. 2008, Marani et al. 2010, Wang and Temmerman 2013, ultimately also resulting in channel formation (Fagherazzi and Sun 2004, Temmerman et al. 2007, Schwarz et al. 2018. However, as these feedback loops only occur after a critical biomass/density has been exceeded (Bouma et al. 2009a), individual marsh propagules such as seedlings or clonal fragments face establishment barriers (Bouma et al. 2009b, 2016, Balke et al. 2014, 2016, Hu et al. 2015, Yuan et al. 2020).
The problematic establishment of early-stage establishment on tidal mudflats has both experimentally and theoretically been related to require (1) inundation limits of marsh plants (Wang and Temmerman 2013, Balke et al. 2016, van Belzen et al. 2017) and (2) episodic occurring Windows of Opportunity with physical calm periods (Balke et al. 2014, Hu et al. 2015, Yuan et al. 2020. Although both experimental and theoretical approaches demonstrate that abrupt marsh recovery may occur at an elevation above which marsh vegetation can start to grow (Kirwan and Guntenspergen 2012, Voss et al. 2013, van Belzen et al. 2017, haphazard field observations suggest that marsh establishment can be significantly hampered by other factors. That is, in some muddy areas no recovery is observed over prolonged periods, even though their elevations are considered to be sufficiently high for mature marshes to occur. For instance, in Paardenschor (Scheldt estuary, Belgium), a large tidal flat remain unvegetated for several years despite its high elevation and the presence of old marshes in the surroundings (Fig. 1a, b). The rare vegetation establishment that does occur in such "halted" systems is usually linked to the presence of topography relief at (shallow) tidal channel edges (Fig. 1). The latter suggests that besides suitable hydrological inundation regime and calm windows of opportunity, also the presence of topographic heterogeneity to facilitate drainage is needed for early-stage seedlings establishment.
Drainage channels constitute basic pathways for the exchange of water during tidal cycles (Perillo et al. 2019, Schwarz et al. 2014. As the tidal area becomes exposed during ebb, water is drained toward the channels predominantly by surface sheet flow (Fagherazzi et al. 2013). Subsequently, a pressure head develops in the sediments, which depending on the sediment type, may affect soil drainage Sun 2004, Fagherazzi et al. 2004). That is, inundation (during flood) and drainage (during ebb) of intertidal areas in general are a complex process that involves not only the water column above the sediment, but may also affect interstitial water and even groundwater exchange (Winterwerp andKesteren 2004, Perillo et al. 2019 Fivash et al. 2020) and thereby relieve sulfide toxicity to seedling roots (Redelstein et al. 2018) that lack aerenchyma (Burdick andMendelssohn 1987, Jung et al. 2008). Drainage effects are generally observed in coarse sand or peaty sediments (see, e.g., Mendelssohn and Seneca [1980] for peat soil; and Padgett et al. [1998], Padgett and Brown [1999] in artificial sandy sediments), where hydraulic resistance of the sediment is typically low and capillary forces in the wide pores are restricted, causing interstitial water to drain vertically from the sediment (Winterwerp and Kesteren 2004). However, vertical drainage is strongly restricted in fine clay (or silt-rich) sediments and may even not be noticeable at all, because hydraulic conductivity of sediments decreases non-linearly with decreasing particle size, and capillary forces become stronger as pore sizes decrease (Winterwerp andKesteren 2004, Hill et al. 2013, Ren and Santamarina 2018, and references therein). Therefore, it is necessary to experimentally test if in such muddy systems with poor hydraulic conductivity of the sediment, surface drainage (runoff of an overlaying water), as may occur due to a topography relief, may offer young seedlings an improved establishment chance, while not having a major effect on the performance of larger marsh plants with well-developed aerenchyma.
In this study, we use an experimental approach to test (1) if poor surface drainage, in a theoretically undrainable, recently deposited and poorly consolidated muddy sediment, forms a barrier for initial seedling establishment, while not affecting seedlings at a larger stage, and (2) if the strength of this barrier depends on the level of elevation relative to mean sea level. We focus on Spartina anglica, which is a typical, widely distributed pioneer marsh species (Gray et al. 1991,  v www.esajournals.org 3 September 2021 v Volume 12(9) v Article e03755 Strong andAyres 2013, Cao et al. 2018). Surface drainage and elevation gradient were manipulated in situ by placing a set of field mesocosms, referred to as "Mega-Marsh Organs" (MMOs), with contrasting surface drainage conditions along the elevation gradient of a sheltered tidal flat. By our MMO designs, we aim to mimic the presence vs. absence of surface water drainage as can be observed at small channel banks in muddy tidal systems. In the first year, seedlings of three successive age classes (i.e., 1-week, 3month, and 1-yr seedlings) were transplanted into the MMOs to quantify the effects of surface drainage, seedling age, and elevation relative to mean sea level on their chances of establishment.
In the second year, the MMOs were used to study the growth performance of mature S. anglica tussocks. Overall, the present study provides insight in the early establishment processes in salt marshes, thereby supporting the development of measures to enhance successful marsh (re)creation, as needed for, for example, sustainable coastal defense.

Study site
Our field experiment was set up at Perkpolder (midpoint: 51°23 0 27″ N, 4°1 0 25″ E), which is located in the Western Scheldt Estuary in the SW Netherlands (Fig. 2a). The estuary is semidiurnal, with tidal range varies from 440 to 550 cm (Baeyens et al. 1997). Tidal flat of this estuary has a mild bed slop (ca. 3‰), with dominated sediment type of mud (Kuijper et al. 2004). The pioneer vegetation in the estuary consists mainly of common cordgrass, Spartina anglica, which has forming monoculture marshes in the seaward part with elevations ranging from 60 to 200 cm NAP (Normal Amsterdam Peil, which is Dutch Ordnance Level that approximately equal to mean high water level in the Scheldt estuary) (van de Wal et al. 2008).
Perkpolder was formerly embanked agricultural land (polder), which was converted to an intertidal zone in June 2015 for nature restoration purposes, by breaching the dike and reintroducing tides from the Western Scheldt Estuary (Rijkswaterstaat 2015) (Fig. 2b). Following the breach, a soft mud layer of around 0.2 to 0.3 m thickness was quickly evolved on top of the highly compacted former agricultural land. After 6-8 months of the breaching, the area stopped morphodynamics and reached equilibrium (Brunetta et al. 2019). Measurements from 2016 to 2018 indicated that the bed-level elevation of tidal flat in Perkpolder had slightly increased to almost nothing (van de Lageweg et al. 2019). The elevation of tidal flat in Perkpolder ranged between −80 and+110 cm NAP (compare mean low water at Perkpolder: −2.06 m NAP; mean high water at Perkpolder: +2.56 m NAP), and therefore, the whole area was completely covered by water and completely dried every semi-diurnal tide cycle from the breach (Brunetta et al. 2019, van de Lageweg et al. 2019. The wave actions in Perkpolder were negligible, as it was almost completely surrounded by seawalls (van de Lageweg et al. 2019) (Fig. 2b). As of the end of our experiment (i.e., fall 2018), natural vegetation had still not established on the newly created intertidal flats, and the deposited sediment still consisted of soft, water-saturated muddy sediment (with a D 50 of 27.62 μm, silt content T 63 of 62.68%, classification c.f. Shepard, 1954). Perkpolder was therefore an ideal site for studying the factors affecting survival of Spartina seedlings/plants of different ages: surface drainage of a theoretically undrainable, recently deposited and poorly consolidated muddy sediment in interaction with elevation to mean sea level.

Marsh establishment experiment-Mega-Marsh Organs with contrasting surface drainage
Mega-Marsh Organs (MMOs; Fig. 3) were used as field mesocosms to study the establishment of seedlings. The original marsh organ concept was introduced by Morris (2007), Kirwan and Guntenspergen (2012), and Voss et al. (2013), in which narrow PVC pipes with drained bottoms were used to test plant growth response to inundation in relation to sea level. To identify the potential effects of surface drainage on seedling survival, we modified and scaled up the marsh organ to allow for hosting multiple plants and surface drainage treatments in the field. In our experiment, each MMO consisted of two large adjacent boxes (inner dimensions of 75 × 45 × 30 cm each, Fig. 3 a, b), both with a closed bottom. In simulation of surface drainage in muddy systems (as might occur due to topography relief at channel edges, see introduction), we equipped one of the two v www.esajournals.org 4 September 2021 v Volume 12(9) v Article e03755 v www.esajournals.org 5 September 2021 v Volume 12(9) v Article e03755 boxes in each MMO with a surface drainage system ( Fig. 3a, b), while left the other one untreated. The surface drainage system consisted of a perforated pipe covered with geotextile (pipe inner diameter of 4.5 cm) at each box side and that extended to the rim to facilitate dewatering of the surface water film. These pipes were connected at the base of the box to a pipe leaving the box, to allow for free water outflow after tidal inundation (Fig. 3a). The geotextile covering the drainage pipes allowed water penetration, while prevented for sediment loss. To avoid sediment clogging the outflow of the drainage pipes, a plastic pallet was placed underneath each MMO to elevate the box bottoms 10 cm higher than the local tidal flat. Given the limited vertical drainage nature of fine muddy sediments in our study, we kept all bottom of the boxes in our MMOs, so as to capture the difference of surface water dewatering via drainage pipes between the two boxes in each MMO. Overall, 12 MMOs (i.e., 24 boxes in total, 12 with surface drainage systems and 12 without) were set up on the bare tidal flat in Perkpolder ( Fig. 3c, d). By using a real-time kinematic global positioning system (RTK-GPS) device, we placed the MMOs at four elevations with three replicate MMOs (3 m interval) per elevation (Fig. 3c, d). The four elevations are 0, 30, 60, and 90 cm leveled to the top of MMO rims respect to NAP (Normal Amsterdam Peil, which is Dutch Ordnance Level that approximately equal to mean high water level in the Scheldet estuary). Inundation time (percent time flooded) of these four elevations during the 2016 experiment period were 56.56%, 49.51%, 45.44%, and 41.41% (Table 1), respectively, which were measured with one SENSUS Ultra device (ReefNet) attached to the top rim of a MMO at each elevation. Ten days prior to planting, all boxes were filled with local fine tidal mud (sediment friction see above).

Marsh establishment experiment-seedlings of different ages
To investigate the influence of drainage in relation to elevation on seedling survival for different age classes, we manipulated a seedling establishment experiment in the growth season of 2016. Seedlings of S. anglica, which is a dominant pioneer species in this part of the estuary, were germinated in a climate chamber with an alternating temperature condition (30°C during the day and 25°C during the night to speed up germination process, c.f. Cao et al. 2018). To obtain seedlings, S. anglica seeds were soaked in nature seawater (32 ppt) in a 4°C refrigerator until germination was performed. Freshwater was then used to accelerate germination and to avoid seedlings drying out. All seeds with a visible sprout were identified as seedlings. Seedlings were transplanted to a tray filled with sandy mud in a climate room and watered regularly with a mix of seawater and freshwater until they were transplanted into the MMOs in the field. To be able to compare seedlings of different age classes, three batches of seedlings were germinated, so that they were 1 week, 3 months, and 1 yr of age at the moment of transplantation.
Before transplanting into the MMOs, seedlings were carefully washed out from the sediments in which they had been grown in the laboratory. Care was taken that transplantation into the MMOs always occurred within 12 h. The roots of all seedlings were planted at the depth of 1 cm in the sediment. In order to address the importance of seedling age, we used six replicates for the 1yr seedlings, six replicates for the 3-month seedlings, and 12 replicates for the 1-week seedlings within every experimental treatment group (24 seedlings in total each box). The experiment was set up on July 2016 and lasted for six weeks, during which surface drainage treated MMOs were checked regularly in avoidance of sediment blocking. As we did not want to disturb the survival conditions for the vulnerable seedlings in the MMOs by other measurements, only the survival of seedlings was recorded every week over a six-week period. A 6-week period was chosen as experimental time frame, as initial survival was the most important component for seedling establishment (Balke et al. 2014, 2016, Hu et al. 2015, Cao et al. 2018). The age effects on seedling survival were accounted for by planting seedlings of different age classes, which allows comparing all age classes under similar growing conditions.

Marsh establishment experiment-mature marsh plants
To compare the potential effects of surface drainage and elevation relative to mean sea level on mature marsh plants, we carried out a followup tussock transplant experiment using the same MMOs. On April 2018, S. anglica tussock transplants were extracted from an existing monoculture marsh from Knuitershoek (close to Perkpolder, Fig. 2a), where uniform size soil blocks (20 × 20 × 30 cm) were collected by cutting of all above-ground vegetation. The tussock transplants were then placed in the abovementioned 12 MMOs in Perkpolder (i.e., 12 treated surface drainage and 12 untreated) with two plant donor soil blocks diagonally embedded in each box in avoidance of limitations of expansion space. Regrowth of each tussock was quantified by clipping all above-ground vegetation as close to the sediment surface as possible at harvest in October 2018. When no above-ground biomass was present at this time, the tussock was deemed as establishment failure. After measuring aboveground vegetation height and recording shoot numbers of each tussock, all the collected plant material was dried at 60°C in an oven to a constant weight for measuring dry biomass.

Marsh establishment experiment-Sediment properties
To compare the sediment properties (in terms of water content and bulk density) between the surface drained and untreated MMOs, we sampled three replicate sediment cores with a syringe (inner diameter = 2.8 cm, depth = 3 cm) on the top of sediment in each box of the MMOs at the end of the seedling establishment in 2016 and the mature marsh establishment in 2018. In total, we collected three replicates per elevation from each MMO box for both marsh seedling (2016) and mature marsh (2018) establishment experiments. Each sediment sample was stored in a container and weighed for wet weight. All sediment samples were then freeze-dried for 72 h and weighed again to calculate water content and dry bulk density using the following formula: Water content ¼ wet sample weight À À

Statistical analysis
Cox regression analysis (Cox 1972) was used to investigate the main effects of each variable (i.e., surface drainage, seedling age, and elevation relative to mean sea level) upon the survival time of seedlings. For which, the death of a seedling was considered as a hazard event during analysis. During the Cox regression, the hazard ratio Exp (B) of each variable was calculated relative to its baseline of each variable, that is, no surface drainage, 1-week seedling age, and 0 cm elevation relative to mean sea level separately, applying a confidence interval of 95%. In general, the Exp(B) was interpreted as a relative risk of the treatment group compared to the baseline (control or placebo group), which was calculated as the ratio of hazards between individuals whose values of hazard differ by one unit when all other covariates were held constant (Cox 1972). Three-way ANOVAs were also applied to test the interaction effects of surface drainage, seedling age, elevation relative to mean sea level upon the survival time of seedlings. Two-way ANOVAs were used to test the effect of surface drainage treatment and elevation relative to mean sea level on tussock survival, plant height, shoot numbers, dry biomass, and sediment properties. The significance level of 5% was used for all analyses. All analyses were performed with SPSS 18.0 software (SPSS, Chicago, Illinois, USA).

Seedlings establishment
Both Cox regression and three-way ANOVAs results revealed that the surface drainage treatment significantly affected S. anglica seedling survival during the 6-week experiment with the poorly consolidated mud of Perkpolder ( Fig. 4; P < 0.05, Table 2). Compared to the groups that untreated with surface drainage, surface drainage treatment reduced the hazard ratio (relative to baseline of surface drainage untreated groups) to (Exp(B) = 0.721 (%95 CI: 0.591-0.880, Table 2). Notably, for 1-yr seedlings at 90 cm elevation, there was significant more (P = 0.036, ttest) seedling survived (94 AE 6%) at week 6 in MMOs that were treated with surface drainage as compared with 67 AE 9% where it was not (Fig. 4). For 3-month seedlings at 30 cm elevation, the survival rate at week 6 was 72.2 AE 15% in MMOs with surface drainage, which was significantly higher (P = 0.014, t-test) than the seedling survival rate 17 AE 9% in MMOs untreated with surface drainage at the same elevation (Fig. 4).
Moreover, the overall survival of S. anglica seedlings (Fig. 4) was significantly affected by seedling age (P < 0.001, Tables 2, 3) and elevation relative to mean sea level (P < 0.01, Tables 2, 3). For S. anglica seedlings of all elevation differing in age, the survival of both 3-month and 1-yr seedlings were all significantly higher in compared to 1-week seedling ( Fig. 4; P < 0.001, Table 2), with a decrease of hazard ratio (relative to baseline group of 1-week seedlings) as seedling age increases: from 3-month seedlings (Exp (B) = 0.230, %95 CI: 0.177-0.299, Table 2) to 1-yr seedlings (Exp(B) = 0.095, %95 CI: 0.067-0.135, Table 2). The steepest decrease in S. anglica seedling survival has been found directly after transplanting (i.e., from week 1 to week 2) for the 1-week-old seedlings, regardless of drainage v www.esajournals.org 8 September 2021 v Volume 12(9) v Article e03755 Fig. 4. Spartina anglica seedling survival during the 6-week field experiment in Perkpolder in 2016. The overall survival time of seedlings in surface drainage treatment (orange) was significantly higher than in the untreated groups (blue). Increased seedling age and elevation relative to mean sea level (from 0to 90 cm NAP) also facilitated seedling survival (data are presented as mean values + SE). Notes: Surface drainage treatment, seedling age, and elevation (relative to mean sea level) were set as variables to analysis their overall effects on seedling survival.
† Exp(B) indicated for the ratio of hazard risk relative to the baseline of each variable, that is, untreated with surface drainage, 1-week seedling age, and 0 cm elevation in this case. The 95% confidence interval was also presented right to Exp(B). In general, the Exp(B) was interpreted as the relative risk of the treatment group compared to the baseline (control or placebo group), which was calculated as the ratio of hazards between individuals whose values of hazard differ by one unit when all other covariates are held constant (Cox 1972). v www.esajournals.org 9 September 2021 v Volume 12(9) v Article e03755 treatment or elevation (Fig. 4). This meant that the older S. anglica seedlings become, the less vulnerable they were to adverse surfacedrainage-lacking conditions. For S. anglica seedlings of all ages differing in elevation, Cox regression showed that the hazard ratio relative to the baseline of 0 cm NAP (Normal Amsterdam Peil, which is Dutch Ordnance Level that approximately equal to mean high water level in the Scheldet estuary) diminished when the elevation increased from 30 cm NAP (Exp (B) = 0.87, %95 CI: 0.66-1.14 Table 2) to 90 cm NAP (Exp(B) = 0.65, %95 CI: 0.53-0.92, Table 2). It was noted that the hazard ratio (relative to 0 cm elevation) at 60 cm NAP (Exp(B) = 0.70, %95 CI: 0.49-0.90, Table 2) was close to which at 90 cm NAP (Exp[B] = 0.65, %95 CI: 0.53-0.92, Table 2). Compare to 0 cm NAP elevation, the overall seedling survival was significantly enhanced both at 60 cm NAP ( Fig. 4; P = 0.003, Table 2) and 90 cm NAP ( Fig. 4; P = 0.011, Table 2). This indicated that the survival of seedling was significantly enhanced when the elevation increased to above 60 cm NAP. However, three-way ANOVAs did not show interactive effects of surface drainage, seedling age, and elevation relative to mean sea level on the overall survival time of S. anglica seedlings (Table 3).

Mature marsh plants
In contrast with the seedling results, for mature S. anglica tussocks, survival was similar among the surface drainage and elevation treatments (Fig. 5), as no significant effect was found on the survival of tussocks from different surface drainage or elevation treatments (P > 0.05, Table 4). Plant height, shoot numbers, and dry biomass of S. anglica were reduced as elevation decreased ( Fig. 5; P < 0.01 for plant height and dry biomass; P < 0.001 for shoot numbers, Table 4), with no significant effect of surface drainage treatment or interactive effect of surface drainage and elevation relative to mean sea level ( Fig. 5; P > 0.05, Table 4). This meant that mature S. anglica tussocks survived better than seedlings at lower elevations and poorly surface drained conditions.

Sediment properties
We did not observe any effect of surface drainage on the sediment properties in our MMOs. The water content of sediment from the MMOs in both 2016 and 2018 did not show any significant effect from either drainage or elevation ( Fig. 6; P > 0.05, Table 5). Average water content of sediment from both years was around 26%, regardless of surface drainage or not. For example, in 2016, for surface drainage treated MMOs, the water content of sediment ranged from 24 AE 1% at 30 cm NAP to 26 AE 1% at 0 cm NAP; and for the MMOs untreated with surface drainage, the water content of sediment ranged from 25 AE 1% at 30 cm NAP to 27 AE 4% at 0 cm NAP; in 2018, for surface drainage treated MMOs, the water content of sediment ranged from 26 AE 1% at 30 cm NAP to 29 AE 2% at 0 cm NAP; and for the MMOs untreated with surface drainage, the water content of sediment ranged from 27 AE 3% at 90 cm NAP to 28 AE 2% at 0 cm NAP. This was confirmed by the dry bulk density data from both 2016 and 2018 that sediment properties were not affected by either surface drainage or elevation ( Fig. 6; P > 0.05, Table 5), indicating that neither interstitial water drawing, nor sediment compaction was changed by the surface drainage treatment over the course of the experiment.

DISCUSSION
The establishment and persistence of salt marsh are highly relevant for conservation and restoration of coastal salt marsh ecosystems. Yet, our understanding remains limited on the processes controlling salt marsh establishment and persistence, especially on bare tidal flats in muddy areas. As muddy sediments typically experience a restriction of vertical soil drainage, we expect that a surface drainage as might occur due to a topography relief on the edge of tidal channels can facilitate the establishment of salt marsh vegetation on muddy tidal flats. Using well-designed MMOs with surface drainage Fig. 5. Regrowth of Spartina anglica tussocks at harvest in the field experiment in Perkpolder in 2018. No significant effect was found on the survival of tussocks from different drainage (surface drainage treatment groups in orange, untreated groups in blue) or elevation treatments. Plant height, shoot numbers, and dry biomass were only affected by elevation relative to mean sea level (data are presented as mean values + SE). Error bars for the survival at 30 and 90 cm were not presented because the numbers were same for the replicates (SE = 0). * indicated for significance level <0.05. systems to mimic surface water dewatering, we experimentally showcased for a poorly consolidated muddy system, a positive effect of surface drainage on the initial seedling establishment of the widely distributed pioneer salt marsh species S. anglica.

Surface drainage in poorly consolidated muddy systems and its effects on marsh establishment
In our muddy systems, the removal of the overlying water (Fig. 7) was the most decisive difference between the surface water drained and undrained treatments. Measurements of  sediment water content showed that the water content of these sediments was very high (about 30% by weight, Fig. 6) and did not change between surface drainage treated and untreated MMOs when measured during ebb. There was also no evidence for increased compaction (higher dry bulk density) of the sediment over the course of the experiment (Fig. 6). Both observations were in accordance with theoretical expectations that such fine muddy soil cannot lose much interstitial water during a single ebb tide (see references in introduction). This makes present study starkly contrast to earlier works on drainage effects on salt marshes in systems with relatively coarse sediments (for salt marshes in peaty soils see Mendelssohn and Seneca 1980; for salt marshes in sandy soil, see Padgett et al. 1998, Padgett and Brown 1999, Redelstein et al. 2018 that have a high hydraulic conductivity and therefore can reduce water saturation as a consequence of vertical draining of sediment. The grain size (with a D 50 of 27.62 μm, silt content T 63 of 62.68%) of the sediment in our system was too fine, and the hydraulic conductivity consequently too low, to allow for substantial drainage of interstitial water from the sediment during the low water period (Winterwerp andKesteren 2004, Ren andSantamarina 2018). The only difference between the surface drainage treated and untreated experiments that was clear from visual inspections, was the absence of a thin layer of overlying water on top of the sediment during ebb (Fig. 7). Therefore, present study has captured the main effects of surface drainage on marsh establishment in our muddy system. Our experimental results indicate that the removal of surface water can indeed improve the survival of S. anglica seedlings in muddy systems, with additional facilitation effects from an increase of seedling age, and a higher elevation relative to mean sea level. The facilitative effect of surface water drainage on seedling establishment could be benefited from the alleviation of anoxic conditions in low-elevation inundated soils (Rabouille et al. 2003, Fivash et al. 2020, where oxygen availability typically restricted for young seedling with reduced rooting under increased inundation (King et al. 1982, Bouma et al. 2001, Redelstein et al. 2018. Although our soil sampling method made it unable for us to distinguish oxygen content within the soil profile, the most shallow sediment layer in which the youngest 1-week-old seedlings were planted (top 1 cm layer) was expected to be more oxygenated after surface drainage (Fig. 7) that allows for soil top layer aeration (i.e., for Redox potential at different sediment depth in mesocosms and field see Redelstein et al. 2018, for soil oxygen decline profile during tidal cycle see Fivash et al. 2020). Such aerobic alleviation can prevent directly toxicity to seedlings from microbial sulfide formation in anoxic sediments (Linthurst andSeneca 1980, Lamers et al. 2013). In contrast, the effect of surface drainage was much less expressed or even v www.esajournals.org 13 September 2021 v Volume 12(9) v Article e03755 absent in larger 1-yr-old seedlings and wellestablished mature plants. The latter might be explained by the presence of well-developed aerenchyma in both roots and shoots, by which they could transport oxygen into soil, to terrestrialize inundation stress (Burdick andMendelssohn 1987, Jung et al. 2008). The overall effects make that drainage of a thin surface water layer benefit for oxygen supply to the growth of especially young seedlings, while not affecting larger plants (Bouma et al. 2001, Redelstein et al. 2018. A strength of present in situ study was that we included an elevation gradient in our experiment settings, revealed that decreased elevation (=increased inundation time) increased the hazard ratio for seedling survival and reduced tussock growth. Since increase of inundation would lead to decreased marsh production at suboptimal elevations (Morris et al. 2007, Kirwan and Guntenspergen 2012, Voss et al. 2013, our results indicated that the absolute elevation of our study site relative to mean sea level is below optimal for S. anglica establishment. S. anglica seedling survival was found to be significantly enhanced for elevations above 60 cm NAP (Normal Amsterdam Peil, which is Dutch Ordnance Level that approximately equal to mean high water level in the Scheldet estuary). In fact, this matches well with field observations that dominate S. anglica monocultures in the Scheldt estuary generally occur with elevations ranging from 60 to 200 cm NAP at the seaward part (van de Wal et al. 2008). At lower elevations, however, our experiments revealed that surface drainage could substantially enhance S. anglica seedling survival. Hence, present findings suggest that, for primary establishment of young S. angica seedlings on mudflats, insufficient surface drainage could be a major factor besides inundation that prevent marsh from extending to lower levels (Voss et al. 2013).

Surface drainage effects as potential drivers for habitat transition in salt marshes
The present experimental results confirmed the low establishment probability of seedlings in the poorly drained, fine sediments that generally have a high water content. Although the addition of surface drainage treatment did improve the survival of S. anglica seedlings, the overall survival remained low, and moreover, the surface drainage treatment affected only the overlying water but not the water draining of the sediment matrix. At the same time, however, the overall results suggest that the addition of surface drainage that affected the runoff of overlying water, did improve the survival of S. anglica seedlings and on low elevations. This is consistent with field observations showing that sparse vegetation recruitment in these habitats occurs at the banks of the shallow channels developing in the soft mudflat (Fig. 1). Thus, in these muddy systems, the only possibility to transit habitat from the bare mudflat to the vegetated marsh seems to be mediated by physical processes leaded by topography relief of surface drainage as channel presents (Temmerman et al. 2007, Vandenbruwaene et al. 2013, Wang and Temmerman 2013. Therefore, the results of the present study indicate that surface drainage plays an important role as a driving mechanism for the habitat transition between non-vegetated tidal flats and vegetated marshes. Based on our findings and the relevant literature, we propose a conceptual model (Fig. 8) to show how surface drainage may drive habitat transition in salt marsh ecosystems: On a heterogeneous marsh platform that with good surface drainage due to topographic elevation differences (Fig. 8a), local-scale surface drainage of the elevated sediment patches may favor early plant establishment via, for example, slight improvement of soil aeration (Crooks et al. 2002, Silvestri et al. 2005, elevated microtopography (Fivash et al. 2020) or by removing of a thin water film (Fig. 7). These established plants may subsequently locally stabilize and trap sediment (Bouma et al. 2009a), thereby gradually concentrating the tidal flow in between vegetation patches. This may lead to the initiation of channel incision in between vegetation patches (Temmerman et al. 2005). Thus, the heterogeneous topography is strengthened with the establishment and growth of vegetation enhancing channel formation (Fagherazzi and Sun 2004, Temmerman et al. 2005, 2007, Vandenbruwaene et al. 2013, which further stimulates surface drainage through the channels. As such, this development of surface drainage could further facilitate seedling establishment, resulting in a positive feedback loop between plant establishment and surface v www.esajournals.org 14 September 2021 v Volume 12(9) v Article e03755 drainage in muddy systems (Fig. 8a). In this way, the presence of good surface drainage may initiate a habitat transition from bare mudflats to vegetated marsh, with positive feedbacks enhancing heterogeneity and self-organized long-term stability (Schwarz et al. 2018). In contrast, on bare tidal mudflats where heterogeneous microtopography is absent (Fig. 8  b), giving poor drainage conditions, seedling establishment is hampered (this study), most likely due to factors like, for example, a thin water film limiting oxygen penetration (Fivash et al. 2020) in the top soil and/or low sediment stability (Redelstein et al. 2018). Seedling mortality prevents vegetation-induced topography heterogeneity and ultimately channel formation, thereby facilitating the persistent smoothening of the mud flat surface by daily tides (Temmerman et al. 2007, Vandenbruwaene et al. 2013). This keeps sediment on bare tidal flats smooth and soft, favoring the development into a homogeneous intertidal platform with waterlogged sediments (Temmerman et al. 2007) (Fig. 8b). Overall, this feedback loop keeps the tidal flat bare even at elevations where normally establishment of vegetation would be expected. (Fig. 1a; Temmerman 2013, van Belzen et al. 2017).
Despite that the underlying mechanisms relating microtopography to plant performances remain incomplete from our field study (but see Fivash et al. 2020 for process-based laboratory experiments), present findings and the feedback mechanism schematically described in Fig. 8 have important implications for management options in muddy systems. For example, creating artificial surface drainage relief to facilitate natural seedling recruitment or planting of older and larger transplant units could both be an efficient management strategy to facilitate marsh restoration on tidal mudflats at suboptimal elevations. For low elevations tidal mudflats that young salt  Crooks et al. 2002;Silvestri et al. 2005;Redelstein et al. 2018;Fivash et al. 2020), which may stabilize and trap sediment (Bouma et al. 2009a), stimulate the formation of channels and consequently further improves surface drainage as might occur at channel banks (arrows 3 and 4; Fagherazzi and Sun 2004Temmerman et al. 2005Temmerman et al. , 2007; Vandenbruwaene et al. 2013), resulting in self-organized feedbacks. (b) On the right, a flat topography with poorly surface drained mudflat (arrow 5; logically assumed) hampers seedling establishment (arrow 6; Redelstein et al. 2018;Fivash et al. 2020) may lead to soft sediment and flat topography due to tidal smoothening (arrows 7 and 8; Temmerman et al. 2007Temmerman et al. , 2012; this induces a runaway feedback loop that keeps the tidal flat bare. v www.esajournals.org 15 September 2021 v Volume 12(9) v Article e03755 marsh seedlings cannot successfully establish, restoration might be achieved via adopting larger seedlings (e.g., 1-yr seedlings in the present study). More importantly, the size of the system to be restored should guide which way to go. Given that clonal expansion is much slower than expansion via seedling recruitment (Yuan et al. 2020, Zhu et al. 2020, large systems should aim at facilitating natural seedling recruitment by creating topographic heterogeneity, whereas small systems may benefit from planting. Further testing of present hypothesized model would be beneficial both for (1) identifying those conditions where Spartina species are more or less likely to appear as invasive species, which is a major problem at many locations around the world (Nehring and Hesse 2008; and references therein), and (2) for implementing nature-based coastal protection strategies when depending on the creation of new salt marshes , van Slobbe et al. 2013, Temmerman and Kirwan 2015.

CONCLUSION
The present study highlights the importance of surface drainage in assisting salt marsh seedling establishment in muddy mineral systems that have a poor hydraulic conductance. Based on our experimental results, we proposed a conceptual model for further studies to understand how surface drainage-driven feedbacks may be reinforced to induce salt marsh establishment. Future research should explicitly account for surface drainage related when aiming to understand salt marsh establishment on especially muddy tidal flats. The findings that seedling perform better on more heterogeneous topographies while large planting units are not sensitive to this should be used to optimize salt marsh (re)creation designs in order to improve the outcomes in muddy systems.

ACKNOWLEDGMENTS
T. J. Bouma and L. Yuan contributed equally to the work reported here. We thank Daniel Blok, Lennart van IJzerloo, Jeroen van Dalen, and Bert Sinke for their kindly help in setting up the experiment. We also thank Greg Fivash for his help in data processing. This work was supported by the National Key Research and Development Program of China (Grant: 2016YFE0133700); the National Natural Science Foundation of China (Grant: 41876093); the research program BE SAFE, which is financed by the Netherlands Organisation for Scientific Research (NWO) and societal partners (grant # 850.13.011); and the CoE-Perkpolder project, which is financed by RWS and the CoE-Delta technology.