Environmental determinism versus biotic stochasticity in the appearance of plant species in saltmarsh succession

1 Terrestrial Ecology Unit, Department of Biology, Ghent University, K. L. Ledeganckstraat 35, BE-9000 Ghent, Belgium 2 Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, P.O. Box: 46414-338, Noor, Mazanderan Province, Iran 3 URU 420 – University of Rennes I, Campus de Beaulieu, 263 Avenue du Général Leclerc, FR-35042 Rennes Cedex, France 4 Research Institute for Nature & Forest, Department of Biodiversity & Natural Environment, Kliniekstraat 25, BE-1070 Brussels, Belgium *Author for correspondence: maurice.hoffmann@inbo.be

Pl. Ecol. Evol. 143 (1), 2010 (stochastic) factors infl uence it (de Leeuw et al. 1993, Erfanzadeh et al. 2010).Salt-marsh vegetation dynamics can therefore be considered to be quite complex and the spatial and temporal patterns of halophytic species and their interactions with multiple biotic and abiotic factors in their own environment remain largely unknown (Silvestri & Marani 2004).To elucidate the underlying reasons for species distribution patterns in salt marshes, most investigators have relied on correlations between plant distributions and soil properties (e.g.Gray & Bunce 1972).An important inference of this approach is that the set of physical conditions associated with each species represents the preferred or optimal habitat of that species (Vince & Snow 1984).
Chance events or historical factors, such as variation in the weather, disturbance intensity, colonization potential and seed availability are generally considered to be more important in determining spatial patterns and temporal changes in early successional stages, whereas the importance of site characteristics, such as soil factors, usually increases with succession age.This generalisation has been proven valid for several terrestrial habitats (e.g.old fi elds (Myster & Picket 1990, Lepš et al. 2000), forest (Christensen & Peet 1984, Mc-Cune & Allen 1985, McClanaham 1986, Tsuyuzaki 1989) and sagebrush steppe (McLendon & Redente 1990)).Therefore, it can be hypothesized that the relationship between environmental factors and vegetation distribution is stronger in old than in early successional stages.Hence, the predictability of occurrence of a species related to soil factors is expected to be higher in old stages than in initial stages.
To test the validity of this hypothesis of environmental determinism of old successional stages versus stochasticity dominance in early successional stages in salt marsh conditions, we compared the relative importance of abiotic environmental factors on the vegetation composition and on the probability of occurrence of individual species between a young and an old salt marsh.A unique opportunity to test the hypothesis in salt marsh conditions was created by a large-scale nature restoration project along the Belgian coast, which led to the creation of a new salt marsh adjacent to an old one (Hoffmann 2006a, Erfanzadeh 2009).Earlier research on critical success factors in the process of man-induced salt marsh restoration was reported by Wolters et al. (2005bWolters et al. ( , 2008) ) for the Blackwater Estuary in south-east England and for comparable initiatives all over north-west Europe in general by Wolters et al. (2005a).A comparison to test the "early stochasticity -late determinism hypothesis" has not been performed though in salt marsh conditions.Expected deterministic abiotic environmental factors considered in the present study were soil salinity (estimated indirectly through electric conductivity (EC) of soil water), sediment texture and elevation (the latter as a surrogate factor for inundation frequency; Hoffmann et al. 2006a).
If the above-mentioned "early stochasticity -late environmental determinism hypothesis" would be correct, we expect that there is a lower correlation between abiotic environmental factors and plant species distribution patterns in new salt marsh conditions than in old salt marsh conditions, since in the fi rst colonization potential and stochastic factors are expected to be dominant over deterministic environmental differentiation.The reverse is expected in the latter, where time has allowed all locations to be potentially colonized by all halophyte diaspores present in the area.Following the early stochasticity -late determinism-hypothesis, we therefore expect species composition and plant occurrence to be more determined by abiotic factors (i.e.salinity, elevation and soil texture) in the old marsh than in the new one.A prerequisite to test the validity of the hypothesis is that no dispersal limitation from the adjacent old marsh to the newly created marsh exists (Wolters et al. 2005b).Knowing that both adjacent marsh areas have a long and immediate contact zone with a low barrier at an elevation level below mean spring tide height, it is expected that indeed no hydrochoric seed dispersal limitations exist between both areas (for a situation map, see Erfanzadeh 2009).

MATERIAL AND METHODS
Nomenclature of taxa follows Lambinon et al. (1998).

Study area and vegetation
The study was carried out in the estuarine part of the Flemish Nature Reserve the IJzermonding, situated along the Belgian coast.After a large-scale nature restoration project, in which approx.half a million m³ of sand and slurry was removed from the original salt marsh -sand dune substrate between 1999 and 2003, primary succession on the newly created intertidal area (approx.12 ha) started in 2002, when part of the newly created marsh was already suitable for colonization (Hoffmann et al. 2006a).In the newly created intertidal area a full inundation frequency ecotone was created (from 0 to 100% inundation of all high waters).The substrate texture was very heterogeneous, ranging from relatively coarse sandy texture (dominant) to local clayey texture.An initial seed bank analysis proved that the new salt marsh substrate was devoid of plant seeds (Stichelmans 2002, cit. in Hoffmann & Stichelmans 2006).During our survey, three years after plant colonization started, dominant plant species on the new marsh were Salicornia europaea (mean total cover in all relevés approx.12%) and Suaeda maritima (11%).In general, annual plant species were largely dominant, perennial plant species were still rather infrequent and very low in cover (always < 1%).
Immediately adjacent to the developing salt marsh, an old salt marsh of approx.5 ha, with full 0-100% inundation frequency range and variable soil texture (comparable to the soil texture variation in the new marsh), remained intact for at least two hundred years (the earliest cartographic evidence dates from 1779 on the De Ferraris maps).Dominant plant species were Elymus athericus (average cover 29,5%, mean characteristic cover 49,2%), Limonium vulgare (6,1% and 14,6%, respectively) and Puccinellia maritima (6,7% and 15,2%, respectively), accompanying but never dominating plant species were Aster tripolium, Festuca rubra and several other salt tolerant herbaceous annual and perennial plant species; in general, perennial plant species dominated over annual plant species; none of both salt marsh areas had been grazed in recent decades until the survey was carried out; nature management was restricted to local measures, such as debris removal (Hoffmann 2006b).

Vegetation analysis
Between mid August and late September 2005, vegetation was studied in a large set of relevés, each 2 × 2 m.In total, 95 relevés were sampled in the newly created salt marsh and sixty relevés in the old salt marsh.Plant species cover was estimated using the decimal scale of Londo (1976).
The elevation of relevés was measured using a 'total station' (Leica TC1600).The reference used is the Belgian Lambert 1972 projection for x-y and the 'Tweede Algemene Waterpassing' (TAW) for elevation.Permanent altitudinal and latitudinal fi xed points from the NGI (National Geographical Institute) network, present in the immediate vicinity of the study area were used as reference points for height and latitudinal measurements.Plot elevation was compared with high tide levels during the ten year period, immediately preceding the nature restoration measures (Fremout 2002).The plots, both in old and new salt marsh, showed an inundation frequency range of 0-100% of all high tides.

Soil analysis
In each relevé three soil samples of the upper 5 cm layer were collected.The three soil samples were pooled per plot and transferred to the laboratory for chemical analyses.After shaking 5 gram of soil in 50 mm distilled water for two hours, the electric conductivity (EC) of the effl uent was measured with a WTW Inolab EC metre level 1. Soil texture was determined with a Coulter LS Particle Size Analyser.Initially, mean, median, and percentages of particle diameter classes of more than 16 and 63 μm of each sample were determined.As the correlation among these particle percentages was very high and previous studies used the median particle size in their research (Jigorel 2000, Langlois et al. 2003), we also used the median particle size as soil texture characteristic.

Data analysis
To test whether a unimodal or a linear response curve should be expected, we fi rst carried out a Detrended Correspondence Analysis (DCA) with Hill's scaling, for both datasets separately.As the length of the gradient was more than 5 in both new and old salt marshes (6.26 and 5.09, respectively), unimodal constrained methods could further be used.To test  which environmental variables would be determining species composition most, we used a Canonical Correspondence Analysis (CCA), again for both salt marshes separately.It enables an evaluation of the relative infl uence of the environment variables on the composition of the community and provides a distribution-free Monte Carlo test of signifi cance (ter Braak & Smilauer 1998).Analyses were done, using CANO-CO for Windows, with 499 permutations.Conditional and marginal effects were estimated (following Lepš & Smilauer 2003), which enables attributing the explained variability to particular variables.Both edaphic factors and elevation level were introduced as explanatory variables.
The probability of occurrence (absence-presence) of the most common salt-marsh species (present in at least 40% of the relevés of one of both salt-marsh areas), i.e.Salicornia europaea, Suaeda maritima, Spergularia sp., Limonium vulgare, Puccinellia maritima and Elymus athericus, was estimated against the environmental variables, using forward stepwise binary logistic regressions (Hosmer & Lemeshow 2000).Spearman's coeffi cient was used to test the correlation between presumed independent environmental variables.As this correlation was highly signifi cant (table 1), the environmental variables were introduced in a PCA and the scores of PCA1, PCA2 and PCA3 were used as indirect, independent variables.

Composition and environmental factors
Species-environment correlation on the fi rst axis in CCA were high (0.92 for the new and 0.85 for the old salt marsh, respectively), indicating that the three environmental variables are potentially important determinants of species variation in both new and old salt marsh.The percentage of spe-Pl.Ecol.Evol.143 (1), 2010 cies variance explained by the three environmental variables together was 17.5% in the old salt marsh and 9.5% in the new one.Environmental variables had a signifi cant effect on species composition in both old and new salt marshes (table 2; fi gs 1 & 2).Nonetheless, the effect of electric conductivity (EC, used as a proxy of salinity) was not signifi cant in the old salt marsh, contrary to the new salt marsh (table 2).

Logistic regressions
From the correlation coeffi cient between PCA axes and the measured environmental variables (table 3), we can conclude that for both young and old salt marshes, PCA1 can be considered as an indicator of soil salinity and to a lesser extent texture, while PCA2 and PCA3 can be considered as indicators of texture and elevation, respectively.
At the species level, environmental variables affected the probability of occurrence in both old and new salt marshes at different levels.In the new salt marsh, the occurrence of Salicornia europaea, Elymus athericus and Suaeda maritima were signifi cantly affected by PCA1 (table 4).The occurrence of Spergularia marina was signifi cantly affected by PCA2 and to a lesser extent by PCA1.The occurrence of Limonium vulgare was signifi cantly affected by PCA3.There was no signifi cant correlation between the occurrence of Puccinellia maritima and any of the PCA axes.
In the old salt marsh, there was no signifi cant correlation between the occurrences of any of the annual species and the PCA axes.There was a signifi cant correlation between the occurrence of Elymus athericus and Puccinellia maritima and PCA3.The occurrence of Limonium vulgare was affected by PCA2 (table 4).

Environmental factors and species composition
Salinity and sediment texture were strongly correlated with elevation.Elevation largely determines the frequency and duration of tidal fl ooding (Packham & Willis 1997), not in the least so in the study areas, where no strong differentiation in levees and backlands appears (Hoffmann et al. 1996), and hence levee structures do not prevent inundation of backlands signifi cantly.It was proven that in salt marshes with a regular inundation frequency, a range of abiotic factors varies in close association with this tidal inundation, such as salinity (de Leeuw et al. 1991, Rozema & van Diggelen 1991), sediment texture (Othman 1980, Thomson et al. 1991), immersion duration (Ranwell et al. 1964), soil redox potential (Armstrong et al. 1985, Groenendijk et al. 1987), disturbance in the form of burial by debris (Brewer et al. 1998) and nutrient levels (Levine et al. 1998, van Wijnen & Bakker 1999).
At the vegetation level, it appears that in both our saltmarsh areas, deterministic factors are important in both early and late successional stages.This is confi rmed by a highly signifi cant correlation between species composition and abio tic conditions in both new and old salt marsh and the signifi cant effect of abiotic conditions on occurrences of most  persal by tidal currents can occur in both new and old salt marshes.In salt-marsh habitats, the effect of stochastic factors (e.g.seed availability) in distribution and patterns of species is thus less pronounced than in terrestrial habitats in the early successional stages.

Determinism of plant occurrence in newly created salt marsh
The composition and distribution of plant communities along the elevation gradient of a salt marsh is related to the ability of individual species to tolerate environmental conditions.In the new salt marsh, the distribution of almost all individual species was affected by and distributed according to abiotic factors and salinity might be the most important of these factors.Indeed, Huckle et al. (2000) showed that salinity and water logging had highly signifi cant effects on growth of some salt-marsh species.Silvestri et al. (2005) stated that at lower elevations, soil water salinity tends to increase due to progressively more frequent fl ooding of the marsh and associated enhanced salt inputs.Consequently, the dependence of soil salinity on elevation may thus partly explain the distribution of species along the elevation gradient.Brereton (1971) proved that salinity and water-logging affected both the distribution of species in a pioneer marsh dominated by Salicornia europaea and Puccinellia maritima.
In our study, Salicornia europaea had the highest positive correlation with electric conductivity.This confi rms fi ndings of Ungar (1998), who stated that this species was the angiosperm that was able to grow at the most extreme end of the salinity gradient.The quick replacement of Salicornia europaea by Elymus athericus in the new salt marsh (as expected by Hoffmann et al. 2006a), because in the old salt marsh Elymus athericus occurs at the same elevation as Salicornia europaea in the new salt marsh, might therefore still take some time.Additionally, Elymus athericus has no easily dispersing diaspores available (Erfanzadeh et al. 2010).Also, Olff et al. (1997)  species in old and new salt marsh.This is in contrast with previous studies in terrestrial habitats.Lepš et al. (2000), studying early succession after top soil removal in a semi-natural spruce forest in an area affected by acidic air pollution, stated that environmental determination of plant community composition increased with successional age.They alleged that the relationship between species and the most important environmental factor, soil texture, was non-signifi cant in the early years but signifi cant in later years and species composition in the early stages of succession was mainly infl uenced by stochastic factors, in their case seed availability.Similar succession observations were made after disturbances in other habitats such as forest (McClanahan 1986, Tsuyuzaki 1989), sagebrush steppe (McLendon & Redente 1990) and abandoned old fi elds (Myster & Pickett 1990, Osbornova et al. 1990, Lepš & Rejmánek 1991).They showed that the avai lability of propagules is usually more important in early stages of succession while site characteristics are more important in late successional stages.Contrarily, in salt-marsh habitats with a daily inundation, propagules disperse very quickly (if the propagules are present).The importance of tidal water in seed dispersal of different salt marsh species between different elevations is well known (e.g.Huiskes et al. 1995, Wolters et al. 2005b, Chang et al. 2007) Ellenberg et al. 1991).Suaeda maritima was comparatively less affected by salinity than Salicornia europaea.Silvestri & Marani (2004) also stated that growth of Suaeda maritima was stimulated at lower salinity levels then Salicornia europaea; the latter reaching maximum biomass production and growth in more saline conditions than the fi rst.In the case of Suaeda maritima dominated communities, it is more likely that in due time, these will be invaded more and more and in the end dominated by Elymus athericus.The latter indeed shows quite similar environmental conditions to those of Suaeda maritima (fi g. 1) and is the more competitive species of both.

Determinism of plant occurrence in old salt marsh
In the old salt marsh, the distribution of perennial species was related to environmental conditions but the distribution of annual species must be affected by other, more stochastic factors.It seems that biotic factors, particularly competition with perennial species, prevented the distribution of these annuals, according to their favoured environmental conditions.Annual species in old salt marshes might germinate in the stochastically appearing vegetation gaps, unrelated to the abiotic factors, where competition with perennials did not prevent the colonization.Indeed, seed germination, cover and biomass of annuals were shown to increase after removal of perennials (Elisson 1987, Gray & Scott 1977) and in the absence of perennials (Packham & Willis 1977, Tessier et al. 2000).Bockelmann et al. (2002) stressed that plant distribution in old salt marshes is a consequence of various interacting factors of which abiotic factors are only a part.Biotic factors like competition between species, local diaspore availability and local habitat adaptation within species are equally crucial for annual species.The fate of annual seedlings largely depends on the permeability of the perennial canopy for light.Thus, disturbance resulting in bare patches within the perennial vegetation seemed essential for the development of annual species (Tessier et al. 2000).

CONCLUSION
From our comparative study of two adjacent salt marsh areas with different age, we conclude that the generally accepted hypothesis that plant distribution patterns in early successional stages is primarily stochastic in nature while these patterns are much more deterministic in late successional stages, is false in a salt marsh environment.Vegetation analysis indicates that diaspores of the annual colonizing species are present ad libitum in the newly created marsh area, but that abiotic factors determine whether they can germinate and establish a local population.Environmental rather than biotic determination of appearance also holds for perennials in old salt-marsh conditions, but not so for annual species, of which it is reasonable to assume that diaspores are equally present ad libitum in the old salt marsh.An interesting next step in understanding vegetation succession in salt-marsh conditions is the development of models that predict the occurrence of annuals in a context of perennial dominance.
As far as expectations on vegetation development in saltmarsh restoration sites is concerned, it seems that further succession beyond a pioneer stage of Salicornia and/or Suaeda maritima dominated vegetation will depend on all three environmental variables measured in this study.The most silty, saline and frequently inundated sites might continue to be dominated by Salicornia spp.for a relatively long time; the more sandy, less saline and more elevated sites, initially dominated by Suaeda maritima, might quickly evolve into Elymus athericus dominated vegetation.The speed of this succession sere entirely depends on the diaspore dispersal capacity of the late successional species and of the evolution of the textural quality of the accreting sediments.Whether the Elymus-stage in succession would further evolve into a Phragmites australis dominated facies, as is the case in more brackish salt-marsh environments such as the mid parts of the Scheldt-estuary (Hoffmann 1993), remains an open question.Within the study area, it seems a possible successional climax in the least silty, least saline environments, since at the most elevated sites of the old salt marshes, indeed Phragmites dominates or starts to dominate the ve getation.Once secondary disturbance, like grazing, is introduced, as is the case in the nature reserve the IJzermonding (a fl ock of sheep was introduced in 2006), succession pathways might change signifi cantly, with more probable dominance of grazing-tolerant grassland species, like Puccinellia maritima and other grass species (see e.g.Olff et al. 1997).
The second author has a postdoctoral scholarship of the Fund for Scientifi c Research (Flanders).
After an interesting and fruitful period of intense cooperation and interaction on this and other manuscripts, our colleague and friend Jean-Pierre Maelfait unexpectedly and extremely regrettably suddenly deceased in February 2009.We wish to dedicate the present paper to his personality and to his contagious devotion to ecological and evolutionary science and, more particularly, his dedicated support of the restoration program of the IJzermonding nature reserve.

Table 3 -Correlation coeffi cient between PCA-components and environmental factors in the old and new salt marsh.
conclude in their study of salt marsh

Table 4 -Variables in the equation calculated by logistic regression. Only
signifi cant variables are shown per individual species.
(van Wijnen et al. 1997)cus only became dominant in the late successional stages.Nonetheless, the newly created salt marsh in the IJzermonding nature reserve, is generally rather coarse in soil texture.Despite the positive relation between clay layer thickness and Elymus appearance and dominance found byOlff et al. (1997)andvan Wijnen et al. (1997)in old, grazed salt marsh systems on a barrier island in the Wadden sea area, we expect the coarse sediment structure at the study site to have a positive effect on Elymus athericus dominance after colonization has taken place.This is indicated by the strong relation between Elymus athericus and PCA1, and between PCA1 of the new salt marsh and texture (see table3 & 4).The species is particularly successful on coarser salt marsh textures (such as levees), since these allow relatively good soil aeration and temporary and quick top soil desalination.Except diaspore availability, soil texture and soil aeration, other factors, such as grazing(van Wijnen et al. 1997), nutrient availability and competition (Bockelmann & Neuhaus 1999) do interfere with Elymus success, making it particularly diffi cult to predict Elymus colonization and dominance in time reliably.When interpreting the salinity data, we at least expect that the salinity level will prevent early expansion of Elymus athericus into the most silty and most saline parts of the Salicornia europaea zone as long as the elevation level does not increase substantially.The two species indeed show a pronounced difference in Ellenberg's salinity index (7 for Elymus athericus versus 10 for Salicornia europaea;