Spatial and temporal patterns of initial plant establishment in salt marsh communities

Questions: 
How are dispersal processes, abiotic and biotic interactions determining the initial salt marsh plant community establishment and development when connectivity is different? We aim to answer this question by analysing the spatial and temporal patterns of plant establishment along the environmental gradient at two connectivity settings. 
 
Location: 
Back‐barrier salt marsh and tidal flats of Spiekeroog, northwest Germany 
 
Methods: 
We established an experiment along the salt marsh elevation gradient with bare sediment open for spontaneous colonisation on the natural salt marsh and on the experimental salt marsh islands built on the tidal flats approximately 500 m from the natural salt marsh for low connectivity. Plant establishment was identified from georeferenced photos at least monthly. 
 
Results: 
Experimental islands as low connectivity plots had limited colonisation by annual halophytes that produce large number of small seeds. Number of individuals increased with higher connectivity at salt marsh enclosed patches. Number of individuals was highest at the mid elevations whereas peak species richness was at the upper salt marsh. Temporal patterns of seedling emergence showed increasing plant numbers until end of April followed by gradual incline over the season at the pioneer and lower salt marsh zones. Upper elevations on the other hand had a stable number of low individual counts over time. Spatial clustering of plant individuals indicating possible facilitation was important at the initial stages of salt marsh development at pioneer and lower salt marsh elevation, but only early in the season. 
 
Conclusions: 
Stochastic patterns of early salt marsh colonisation indicated that success of species colonisation was determined by seed properties, seed availability and environmental conditions mediated by elevation. We found indications, that further colonisation was supported by already colonised plants at initial stages, but shifted to avoidance later in the season.


| INTRODUC TI ON
The long-term stability and resilience of ecosystems depends partially on the species' ability to recover and regenerate after disturbances, which is realised by colonisation and establishment (Loreau et al., 2003;Leibold et al., 2004). Plant communities in active sedimentary environments such as salt marshes, are subjected to frequent physical disturbance and benefit from this ability to re-colonise disturbed areas (Adam, 2002;Leonardi et al., 2016;Almeida et al., 2017). Salt marshes are globally under pressure from anthropogenic developments and accelerated sea-level rise (Adam, 2002;Gedan et al., 2009;Kirwan et al., 2016). Additionally, frequent physical disturbance due to storm erosion and burial of plants by flotsam and sediment play a crucial role in structuring salt marsh communities (Reed et al., 2018;Schuerch et al., 2018). Establishment, resilience and long-term stability of salt marshes is, therefore, highly dependent on the species' abilities to re-colonise and establish after disturbances at different scales. Understanding the species-specific dispersal and re-colonisation pattern is therefore important to conserve salt marsh biodiversity but also for the provision of regulatory ecosystem services such as coastal protection Möller et al., 2014;Zhu et al., 2019).
Salt marsh communities are predominantly connected through hydrochorous diaspore dispersal with tidal currents (Wolters et al., 2004;Chang et al., 2007). Seed traits related to hydrochory and long-distance dispersal, such as the number of seeds produced per individual and buoyancy, are thus particularly important in contributing to dispersal success Chang et al., 2016;Raju & Kumar, 2016). Long-distance dispersal can be mediated by shorebirds and geese, especially during migration (Tóth et al., 2016;Lovas-Kiss et al., 2018). Interestingly, most dispersal in salt marshes has previously been described as local with limited long-distance dispersal mainly during storm events (Chang et al., 2007). Although there is evidence for successful long-distance seed transport, the seeds need to be trapped in the salt marsh and deposited at suitable microhabitats, which could be hampered during storm conditions (Chang et al., 2007). Dispersal of plant species by salt water is generally understudied and specific knowledge about the dispersal distances of salt marsh species is limited (Bullock et al., 2017).
Plant community assembly is not only the result of dispersal but also requires successful establishment (Rand, 2000). Colonisation therefore also depends on suitable environmental conditions for the species to germinate and establish (Morzaria-Luna & Zedler, 2007;Wolters et al., 2008). For salt marsh species environmental conditions such as salinity, flooding frequency and duration determine establishment success (Veenklaas et al., 2015;Edge et al., 2019). The combination of environmental conditions and competitive abilities of species leads to a distinct zonation of vegetation along elevational gradients in European salt marshes (Petersen et al., 2014). The pioneer zone in the lowest elevations is dominated by flood-and salt-tolerant species with little competitive ability. Upper salt marsh at the highest elevation on the other hand is dominated by less stress-tolerant species with strong competitive abilities (Pennings et al., 2005;Minden et al., 2012). Understanding the effects of fragmentation and potential dispersal and establishment limitations on species sorting along environmental gradients would further the understanding of the functioning of metacommunities in general (Grainger & Gilbert, 2016).
The first plant individuals can provide shelter from hydrodynamic forcing, shade the substrate and alleviate salt stress for newly establishing individuals, therefore facilitating further colonisation (Engels et al., 2011). Facilitation is more important at the stressful end of the gradient (Maestre et al., 2009). Competition on the other hand is dominating at less stressful conditions at the upper elevations, where strong competitors can outcompete most other species (Crain et al., 2004;Pennings et al., 2005;Wanner et al., 2014).
The role of competition during the initial colonisation process is nonetheless small, as strong competitors tend to be slow colonisers (Tilman, 1994). Thus, competition is expected to become relevant at later stages of community assembly, when annual halophytes are replaced by perennial turfs (Bertness & Shumway, 1993).
Sea-level rise, shifting water currents, and increasing storminess may increase formation of bare sediment patches open for plant colonisation (Feser et al., 2015). We therefore aim to understand plant establishment during initial salt marsh vegetation development. This was done by quantifying the effects of seed availability, environmental conditions and biotic interactions on salt marsh vegetation assembly. We studied these effects within a metacommunity experiment, where we created isolated salt marsh patches by building experimental islands at the tidal Wadden Sea coast and "connected" patches enclosed in the natural salt marsh (Balke et al., 2017). Here we evaluate the spatial and temporal patterns of colonisation along the environmental gradient of the salt marsh. Specifically, we address the following questions: (a) is plant colonisation determined by connectivity and plant dispersal traits; (b) how does the elevational gradient affect initial colonisation; (c) how are the temporal changes in plant numbers affected by the environmental conditions (i.e. elevation as a proxy for flooding and biotic interactions); and (d) does initial colonisation indicate clustering or avoidance of individuals?
With this paper we provide a unique experimental in situ approach to metacommunity research, specifically a detailed insight into initial community assembly. Upper salt marsh at the highest elevations is dominated by Elymus athericus, followed by a more species-rich lower salt marsh at mid elevations with high covers of Atriplex portulacoides and a Spartina anglica-dominated pioneer zone with high frequency of Salicornia europaea at the lowest elevations. Suaeda maritima had its highest frequency at the mid elevations.

| Study area and experimental design
We set up six 2 × 2 m experimental plots at each elevational zone, i.e. upper, lower and pioneer salt marsh within the back-barrier salt marsh (i.e., representing high-connectivity conditions).
These plots were first cleared of vegetation to 30 cm depth, then filled with sediment from the tidal flat and left bare for spontaneous colonisation. Sides of the experimental plots were lined with root cloth to hamper colonisation from lateral growth. An additional six plots were marked as untreated reference. From each 2 × 2 m replicate plot, two subplots of 1 × 1 m were randomly allocated for non-destructive surveys such as vegetation sampling and seedling counts, and two subplots for destructive sampling, for example soil coring.
Experimental islands were built on the tidal flat approximately 500 m from the natural salt marsh to study the colonisation patterns under more isolated conditions (i.e. low connectivity) (Balke et al., 2017). Each of six experimental islands was built with 12 steel cages of 1 × 1 m. Three elevational levels, with four joint cages per level, correspond to the elevations of the salt-marsh-enclosed plots and hence receive the same amount of inundation. Cage heights of 70, 100 and 130 cm correspond to the pioneer, lower and upper saltmarsh plots respectively. Each individual cage was lined with permeable geotextile to sustain the sediment, and polyethylene bags at the lower section to maintain the ground water levels comparable to that in natural salt marshes ( Figure 1). We then filled the cages with sediment from the surrounding tidal flat, leaving the top 10 cm of each cage empty to allow space for natural sedimentation processes.

Island sediment was then left open for spontaneous colonisation.
We used sediment from tidal flats for both low-and high-connectivity plots to ensure the same soil properties and seed bank possibly stored in the sediment. However based on our own samples (data not published) and the literature (Wolters & Bakker, 2002), there is no considerable persistent seedbank of salt marsh species in the tidal-flat sediments. A detailed description of the experimental setup is published in Balke et al. (2017).

| Data collection
We followed plant establishment on experimental plots in the growing season of 2015, after allowing for spontaneous seed transport over the autumn and winter season of 2014/2015. We set up the experiment in August 2014, i.e. before the major autumn seed dispersal time for most salt-marsh species. Species that disperse earlier in the season, e.g. Cochlearia spp., may therefore have had some disadvantages. To record plant numbers on the experimental plots, we photographed the plots twice in April (9.04 and 28.04) and once in each following month until August. The photos were georeferenced in ArcGIS (ESRI) using corners of the marked plots as reference points. In each photo, we marked and identified all seedlings and plant individuals. We focused on the number of individuals on the plot without following the individuals separately. Therefore we can only analyse the changes in individual numbers and not in mortality or new emergence. In addition to individual numbers we sampled the plant cover on reference and experimental plots by frequency analysis in August 2015. We used a 0.9 × 0.9 m counting frame divided Plant trait data were collected in 2015 from 10 healthy fullgrown plant individuals, when seeds were already ripe, but not yet shed. For selected plant individuals, inflorescences with seeds were clipped and stored in a paper bag. We isolated the seeds and measured the air-dried weight of all seeds and counted their number.
Additional measurements for the same seed traits and species was retrieved from the LEDA database (Kleyer et al., 2008). We averaged the measured trait values per species with approximately 50 individuals per species.
Environmental conditions such as salinity and inundation frequency in the natural salt marsh and on experimental islands were monitored by Zielinski et al. (2018). We calculated the inundation duration as hours when a plot was inundated at least to the soil level (proportion of inundated hours in measured time frame) and inundation frequency as number of times the tide reached plot elevation (Appendix S1).

| Data analysis
To analyse the effect of distance from the main seed source on the number of individuals recorded on the experimental plots we fitted an exponential power model using the lm() function in R (R Core Team, ) with log-transformed individual count as response variable and second power of distance to the seed source as explanatory variable.
Only the three most frequent species were numerous enough to be analysed. We used the individual count from the point of time when the species was most frequent -end of April for Salicornia europaea (from here on Salicornia), May for Puccinellia maritima (Puccinellia) and June for Suaeda maritima (Suaeda). Trade-off between the seed mass and seed size was evaluated by reduced major axis regression (Warton et al., 2006) to aid the interpretation of plant colonisation results. Relationships between the species counts at the peak of the growing season in August and environmental parameters were investigated with non-metric multi-dimensional scaling (NMDS) using the metaMDS function in the vegan package (Oksanen et al., ). We used Bray-Curtis distances with random starting configuration and 100 iterations, which resulted in a final two-axes solution with a stress value of 0.1.
We analysed the differences in plant numbers between the elevation zones and sampling times on salt marsh-enclosed plots using a generalised linear mixed model with Poisson distribution in the lme4 package (Bates et al., 2015). We entered number of individuals as response variable, time and elevational zone as fixed effects and plot id as random intercept. We built separate models for Salicornia and

| Effect of distance on species colonisation
Taking into account both high-and low-connectivity treatments, p < 0.001). Distance from the seed source did not explain the variation in individual counts of Puccinellia. Seed mass and seed number of salt marsh plants displayed a trade-off (R 2 = 0.28, p < 0.05).
Salicornia is a species producing the largest number of lightest seeds, with salt marsh grasses like Spartina anglica and Elymus athericus investing in fewer larger seeds (Figure 3).

| Species establishment patterns along the elevational gradient
Abiotic factors as inundation parameters, salinity and nitrogen availability were highly correlated with elevation and their independent effects could therefore not be distinguished (Figure 4). In the following, we therefore use elevation as indication of soil salinity, nutrient availability and inundation regime. salt marsh-enclosed experimental

| Temporal changes in plant numbers
Temporal patterns were only analysed for salt marsh-enclosed plots and differed slightly between the two most frequent species  was either no effect or avoidance of already densely colonised areas (Figure 7).

| Connectivity and role of plant traits
Connectivity of the salt marsh patches affects the initial colonisation in the recovery processes. Limited dispersal to experimental islands already at distances of approximately 500 m (Figure 2) shows that most dispersal in salt marshes takes place at relatively local scales.
This has been shown previously (Chang et al., 2007), but only within salt marshes, not across uninhabitable mudflats. Dispersal and establishment success results from the interplay of connectivity and traits enabling long-distance dispersal. Species with these traits were successful at reaching the low-connectivity sites. The annual halophytes Salicornia europaea and Suaeda maritima were the most numerous colonisers in our study and together with Puccinellia maritima, the only species that managed to colonise the more isolated experimental islands. Similar initial colonisation and domination of annual species has been noted for both restored areas and in open patches in otherwise intact natural salt marsh (Tessier et al., 2002;Hughes et al., 2009;Balke et al., 2017). Salicornia produces large numbers of very light seeds that are adapted to water dispersal ( Figure 3). The seed dimorphism of Salicornia, where small and large seeds show varying germination responses and possibly different dispersal mechanisms, could add to the success (Ameixa et al., 2016;Liu et al., 2018). Suaeda on the other hand produces fewer slightly larger seeds that have high buoyancy (Raju & Kumar, 2016). The high germination rate of Suaeda seeds may further improve its success as an initial coloniser. Both Salicornia and Suaeda seedlings are also known to tolerate high salinity and waterlogging conditions (Tessier et al., 2000). Although seed traits do not support Puccinellia as a good long-distance disperser, it has an ability to emerge from vegetative parts that can float . Clonal growth and establishment from displaced plant fragments is therefore another aspect in community assembly already at the initial stages.

Distance to main seed sources and reduced inundation events
at the upper salt marsh elevations could lead to insufficient seed arrival within the salt marsh. Bare soil, lower soil salinity and inundation frequency suggest favourable conditions for establishment. The interplay between suitable conditions but lack of diaspores is evident from the low seedling numbers but high species diversity in the upper salt marsh zones. Even when water level reaches higher elevations, the upward transport of propagules from lower salt marsh elevations is generally lower than the seed transport from higher elevations towards the mudflats (Huiskes et al., 1995;Wolters et al., 2004;Zhu et al., 2014). This also means that at elevations with high inundation frequency, such as the pioneer zone, a large proportion of seeds can be transported out of salt marshes into the uninhabitable mudflats at subsequent tides, again reducing the availability of seeds (Zhu et al., 2014). The highest number of seedlings at lower salt marsh elevations could be the result of the location around the elevation of mean high water. The water's edge there could provide substantial input of seeds via wave-skimming, similar to the deposition of floatsam (Tessier et al., 2002).
The small step in elevation between high and low marsh at our study site might further encourage seeds to be stranded at that location.

| Role of environmental factors
After seed arrival, the seedling's emergence and survival are related to environmental factors such as inundation regime (Pétillon et al., 2010), hydrodynamic conditions (Mateos-Naranjo et al., 2008;Redelstein et al., 2018), sediment stability Cao et al., 2018) and salinity (Shumway & Bertness, 1992;Dethier & Hacker, 2005;Hughes et al., 2012). Lower water storage of the sandy tidal-flat sediments could be causing unsuitable soil moisture and salinity conditions during periods of germination, especially in the upper elevations with higher groundwater levels and infrequent inundation. Salinity on the experimental salt marsh and island plots was higher compared to the natural salt marsh conditions in the initial seed germination period (Appendix S5). Salicornia and Suaeda are better adapted to hypersaline conditions that otherwise often hamper seed germination and seedling survival for most plant species (Shumway & Bertness, 1992;Davy et al., 2001;Tessier et al., 2002 (Balke et al., 2017). Similar better growth of halophytes at higher elevations has been observed in subarctic salt marshes (Snow & Vince, 1984). Open spaces after disturbances often enable pioneer species to grow higher in the tidal frame and should similarly support upper marsh species lower in the tidal frame (Sullivan et al., 2018). However, Elymus athericus, a characteristic species of the upper salt marsh, did not colonise the bare patches at lower elevations. We consider this a clear indication of species sorting due to environmental stress, as shown with transplantation experiments (Bertness & Ellison, 1987;Pennings et al., 2005). We did find species from lower elevations colonising bare patches in the higher marsh where they do not occur in natural salt marshes. We therefore expect that after a first strong colonisation from annual halophytes, they are replaced by more competitive species in later years (Tessier et al., 2002).
The question of dispersal vs establishment on our experimental patches, however, needs additional investigation, as our current results are a combination of variation in recruitment and survival as well as dispersal. Disentangling dispersal and establishment functions is challenging due to the many factors involved and measuring dispersal is notoriously difficult and resource-consuming (Bullock et al., 2006).
However, the need for such dispersal studies is great, considering that knowledge about dispersal abilities of salt marsh species is limited and dispersal by water particularly understudied (Bullock et al., 2017).

| Temporal patterns and biotic interactions
Salicornia achieved its largest density by the end of April and declined steadily throughout the season in pioneer and lower salt marsh zones. A similar temporal pattern of seedling emergence and decline was observed in a natural salt marsh by Jefferies et al. (1981).
Changes in Salicornia at the pioneer zone showed clustering at the initial seedling establishment in April and May -increases in plant numbers were then larger at locations where mean distances be- Similar to Salicornia, clustering of Suaeda seedlings only appeared during the main establishment phase and the decrease in numbers at the lower salt marsh seemed to be density-dependent.
In the upper salt marsh plant numbers remained low but stable throughout the season. For both species, seedling increase by facilitation subsequently led to a decline by competition in lower salt marsh zones. These differences in population dynamics on small spatial and temporal scales highlight the relative importance of seed deposition vs density-dependent seedling emergence and mortality. Initial plant communities were shaped by the seed availability and biotic interactions. The importance of small-scale heterogeneity within an elevational zone will probably take effect at later stages of community development (Deák et al., 2015).
Dispersal and establishment success play a crucial role in shaping salt marsh plant communities. Our results confirm that seed availability and environmental conditions determine the initial community of a salt marsh. These results further emphasise the need of restoration efforts to consider the availability of connected seed sources in the vicinity. We however need to have a better recognition of the interplay between dispersal properties and the traits of salt marsh species to cardinally widen the understanding of salt marsh metacommunities.