Colonization of a newly developing salt marsh: disentangling independent effects of elevation and redox potential on halophytes

Authors


Correspondence author. E-mail: a.davy@uea.ac.uk

Summary

1. Many characteristics of the salt marsh environment covary with elevation. It has therefore proved difficult to determine which environmental characteristics limit the distributions of particular species in the field. Oxygen supply to the rhizosphere may be particularly important, as it is determined by the duration and frequency of flooding.

2. The re-activation of a salt marsh by managed coastal realignment provided an opportunity to investigate the large-scale manipulation of environmental effects on halophyte distribution in a situation where the usual relationships between environmental characteristics, elevation and succession had been partially uncoupled.

3. Most locations sampled lay between mean neap and mean spring tidal levels. As expected, anoxic conditions occurred at lower elevation, redox potential increased generally with elevation and sediments were oxic on the upper parts. However, sediment oxygenation at any given elevation was variable, particularly at intermediate levels in the tidal range. This imperfect correlation between elevation and sediment redox allowed quantification of their independent effects on species distributions using the statistical technique of Hierarchical Partitioning.

4. Effects of elevation and sediment redox potential were distinguishable from each other. Salicornia europaea occurred predominantly at lower elevation but was not influenced by redox potential. Puccinellia maritima favoured low redox potentials independently of elevation. In contrast, Suaeda maritima tolerated a wide range of elevations but was absent from areas with low redox potential. Atriplex portulacoides was apparently more averse to low redox potential than to low elevation. Elytrigia atherica was restricted to both high redox potential and high elevation. Smaller independent effects of sediment depth, salinity, water content, nitrate concentration, shear strength and loss on ignition were apparent for some species.

5.Synthesis. Although much of the elevational zonation of species on salt marshes is mediated by differential tolerance of the consequences of co-linearly varying variables, particularly sediment anoxia and elevation, these variables have independent effects that are quantifiable in the field. Hierarchical Partitioning provides a valuable tool for distinguishing the mechanisms underlying species zonations on environmental gradients, especially where large-scale environmental manipulations have partially decoupled the usual co-linear variation.

Introduction

The ability of halophytes to inhabit coastal salt marshes reflects their tolerance of the physical environment there, particularly tidal inundation and salinity. Species distributions may be strongly modified by both positive and negative interactions with other species (Bertness & Ellison 1987; Pennings & Callaway 1992), but in the early stages of succession, physical limitations would be expected to be pre-eminent. Elevation within the tidal range is a primary determinant of the environmental factors that affect plant distribution and is the physical basis of the distinctive zonation often reported for the distribution of salt marsh halophytes (e.g. Chapman 1960; Zedler et al. 1999; Bockelmann et al. 2002). Increasing elevation as a result of sediment accretion is a key feature of the successional processes by which salt marshes develop (Ranwell 1964; Castellanos, Figueroa & Davy 1994; Figueroa et al. 2003). The elevation of the marsh surface has direct consequences for the frequency and duration of submergence experienced. It also has indirect consequences for other aspects of the environmental conditions. Submergence influences the range of salinities that occur at a particular location and retention of the tidal water within the pore spaces in the sediment reduces gas exchange with the atmosphere. The resulting sediment hypoxia is manifested as low redox potentials (Armstrong et al. 1985; de la Cruz, Hackney & Bhardwaj 1989; Castillo et al. 2000; Anastasiou & Brooks 2003), associated with increased concentrations of potentially toxic, reduced substances (e.g. sulphide, manganese II and iron II). Salt-marsh halophytes differ considerably in their ability to oxygenate their rhizospheres (e.g. by the production of aerenchyma) and to tolerate reducing conditions (Colmer & Flowers 2008).

Sediment water content and redox conditions may, however, be affected by factors other than elevation. These include properties of the sediment itself, including its consolidation and hydraulic conductivity, as well as local drainage conditions, such as the proximity of creeks and pans or the presence of a subsurface aquitard (Crooks et al. 2002). Conversely, increasing elevation has consequences other than waterlogging, such as longer exposure to the atmosphere and light for photosynthesis (Chapman 1960) and less destructive wave energies (Brereton 1971). In a naturally structured salt marsh, the effects of elevation, waterlogging, other environmental variables and the interactions between species are all intercorrelated. Elucidation of the mechanisms that regulate the colonization of intertidal sediments by halophyte species requires both experimental manipulations and statistical procedures capable of distinguishing the different effects.

Recent creation or reactivation of salt marshes in south-east England has been driven by the need to maintain marsh area and by the increasing cost of defending existing coastlines in the face of sea-level rise and climate change (French 2006). Because such sites are often created by the restoration of tidal influence to historical land claims (‘managed coastal realignment’), they also provide large-scale field manipulations in which to investigate halophyte colonization at predetermined elevations. Many sites are initially at elevations below that of current nearby marshes because of shrinkage of the land and concurrent sea level rise, and considerable accretion of sediment is required. However, parts of some sites may be at considerably higher initial elevations than those at which salt marsh succession normally commences on intertidal flats. On these, all elevations are exposed to halophyte colonization simultaneously, rather than in a protracted chronosequence. In general, the colonization of managed realignment salt marshes adjacent to existing marshes is not limited by dispersal (Wolters, Garbutt & Bakker 2005; Mossman et al. 2012). Consequently, underlying physical limitations across the whole elevation gradient can be investigated before interactions between species become important. The absence of an effective drainage system and changes to soil structure during the terrestrial phase may also decouple the relationship between elevation and waterlogging. Even in these systems the physical decoupling will only be partial, but clear separation of the effects of potentially co-linear variables can be achieved statistically using computationally intensive Hierarchical Partitioning methods not previously applied to this problem (Mac Nally 2000).

This work examined halophyte colonization in relation to surface elevation and a range of physico-chemical factors on the Brancaster managed realignment site, Norfolk, 4 years after tidal re-activation. We hypothesized that (i) the distributions of individual species would be explained substantially by the environmental variation; (ii) variation in elevation and sediment redox potential would account for much of the variation in early species colonization; and (iii) that there would be significant independent effects on species distribution of redox potential, of elevation and of other factors.

Materials and methods

Site

Work was carried out at the Brancaster managed realignment (MR) site on the North Norfolk coast, UK (52 58 20.359°N, 00 37 53.473°E). Originally salt marsh, this area had been reclaimed during the 18th century to provide freshwater grazing marsh, but tidal flow was restored to a 7.5-ha area as a managed coastal realignment in 2002. A culvert of six 1-m diameter pipes in a lowered section of sea wall, which is now also overtopped by spring tides, allows relatively uninterrupted tidal flow and does not significantly reduce tidal amplitude. A simple, artificial system of branching, linear creeks that drain to the culvert was excavated prior to re-flooding. The non-halophytic vegetation was killed rapidly by seawater, with the exception of small areas of Juncus acutus L., Phragmites australis (Cav.) ex Steud. and Elytrigia atherica (Link) Kerguélen. Dead biomass was allowed to decompose in situ and was incorporated into the soil. Over the first 4 years, almost all of the halophytes found in the North Norfolk coastal marshes became established, indicating that dispersal was not limiting colonization, although populations were sparse and there was much bare ground (Mossman et al. 2012).

Sampling

A regular (square) sampling grid of 208 points at 15-m intervals was established to cover the entire site in March 2006. At each point, elevation relative to the UK reference mean sea level (Ordnance Datum Newlyn, ODN) was measured using a differential global positioning system (DGPS; Topcon, Newbury, UK), with an accuracy of <2 cm and precision of <1.5 cm. Tide levels were measured on-site over 3 months and compared with recorded standard port data for the same period (Mossman 2007). The levels of mean high water neap (MHWN) and mean high water spring tides were, respectively, 1.71 and 3.24 m.

A 1 × 1-m quadrat was located at each sampling point, and the percentage cover of all species and of remaining bare ground was recorded to the nearest 5% (rare species were assigned a value of 1%).

Sediment depth and substrate shear strength were measured in situ at each sampling point in March 2006 and substrate redox potential in June 2006; water infiltration rate was determined at a randomly selected subsample of 20 of the sampling points in June 2006. Where it was not possible to conduct fieldwork within the same tidal cycle, measurements were taken over two consecutive cycles. Substrate samples (c. 10 cm deep, 5 cm diameter) were collected from each sampling point in March 2006 and stored in sealable bags at 4 °C for laboratory analysis. All samples were taken at low tide.

Substrate analyses

At each sampling location, the depth of marine sediment deposited on the soil surface was measured by probing with a mm scale. Three measurements of sediment shear strength were made with a hand vane (Pilcon, Stevenston, UK; range 0–118 kPa) each within 10 cm of the sample point. Substrate redox potential was measured using a combination redox electrode with an Ag/AgCl reference (BDH Gelplas, VWR, UK) and voltmeter (Hanna Instruments Ltd, Leighton Buzzard, UK) at 5 cm depth, left until a stable reading was achieved (up to 5 min). Values were corrected by adding the potential of the reference electrode (204 mV) with respect to a standard hydrogen electrode. Redox potential was not corrected for pH. Duplicate determinations of water infiltration rate were made at a sub-sample of 20 points, using a method similar to that of Howes et al. (1981): transparent plastic tubes (25 mm diameter) were driven into the substrate to a depth of 70 mm, filled with seawater and the change in level timed over a 30-min interval, with topping-up as necessary.

In the laboratory, duplicate subsamples (c. 5 g of each substrate sample) were oven-dried (16 h at 90 °C) to determine water content and then ignited in a muffle furnace (16 h at 390 °C) to determine organic matter content (loss on ignition). Further subsamples of substrate were extracted with distilled water (1:5 soil:water). The salinity (psu) of the extracts was determined from electrical conductivity (Jenway 470; Barloworld Scientific Ltd., Essex, UK), calibrated against artificial sea water (Tropic Marin, Wartenberg, Germany), and the soil water content was used to calculate the salinity of soil pore water. Soil samples were collected after the equinoctial spring tides. This timing represents a compromise between mid-winter when higher marsh sediments tend to be hyposaline and mid-summer when they are often hypersaline. Ammonium (inline image) and nitrate (inline image) concentrations in the extracts were measured using ion selective electrodes (Scientific Laboratory Supplies Ltd., Nottingham, UK), and water-extractable phosphate (inline image) was determined spectrophotometrically by the molybdenum-blue method.

Data analysis

Data were analysed using SPSS 14.0 (SPSS Inc, Chicago, IL, USA) and R 2.4.1 (Development Core Team, R Foundation for Statistical Computing). Correlation between variables was assessed by Spearman’s rank correlation. Generalized linear models were used to examine the relationship between environmental variables and the abundance or occurrence of the commonest species. Models with Gaussian errors were fitted to data on cover of bare ground and the abundance of the two commonest species, Salicornia europaea agg. L. (including mainly the microspecies S. europaea sensu stricto L. and S. ramosissima Woods (Davy, Bishop & Costa 2001) and Suaeda maritima (L.) Dumort. (hereafter Salicornia and Suaeda respectively). Models with binomial errors were fitted to presence/absence data for Puccinellia maritima (Huds.) Parl., Atriplex portulacoides L. and Elytrigia atherica (hereafter Puccinellia, Atriplex and Elytrigia, respectively). The percentage cover data and some of the independent variables were transformed to give approximately symmetric distributions before analysis. Percentage cover of bare ground was square-root transformed, and percentage cover of Salicornia and Suaeda were fourth-root transformed. The independent variables were elevation and eight sediment characteristics: redox potential, salinity, water content, loss on ignition, fourth-root transformed depth of newly deposited sediment, shear strength, nitrate concentration and phosphate concentration. Sediment infiltration rate was not included because of its restricted scale of sampling. In all cases, the relationships with elevation were nonlinear, and in several cases were unimodal. To enable this to be captured by a single independent variable, a quadratic relationship with elevation was fitted to each species individually, and the position of the maximum calculated. This value was subtracted from the elevation data and the result squared.

Several correlations between environmental variables were significant, so the independent contributions of each were determined using Hierarchical Partitioning (HP), and the significance of each contribution was determined by bootstrapping (Chevan & Sutherland 1991; Mac Nally 2000; Mac Nally & Walsh 2004). The nine independent variables used correspond with the limit above which Hierarchical Partitioning suffers from numerical instabilities (Olea, Mateo-Tomás & de Frutos 2010). Hierarchical Partitioning was used in preference to partial ordination methods, such as Canoco, or nonparametric Mantel tests, because of its unique ability to assess the contributions of relationships with each of the variables, even when the independent variables are strongly intercorrelated.

Results

Environmental variables

The sampling points were between elevations of +2.16 and +3.36 m (Fig. 1), all above the level of MHWN tides. The lower and upper elevations receive c. 516 and 43 inundations per year, respectively, and the modal elevation of +2.65 m c. 298 inundations.

Figure 1.

 The distribution of elevation, relative to UK reference mean sea level (ODN) of 208 sample points on a regular, square grid (with 15-m intervals) covering all of the Brancaster managed realignment salt marsh, Norfolk, in 2006. MHWN, mean high water neap level; MHWS, mean high water spring level.

With the exception of four points at high elevation, a layer of freshly deposited sediment was present at every sampling point (Table 1). However, this layer was thin across most of the site (≤20 mm at 86% of points; ≤5 mm at 43% of points). Infiltration rates were generally low, with the majority <0.07 mm s−1 (= 20).

Table 1.   Summary of environmental variables measured at Brancaster managed realignment salt marsh, 2006
VariableMedianConfidence interval (95%)Rangen
Elevation (m)2.642.34–3.042.16–3.36208
Sediment redox potential (mV)311−79 to +379−201 to +509207
Sediment infiltration rate (mm s−1)0.90.1–13.00–0.520
Superficial sediment depth (mm)80–600–110207
Sediment shear strength (kPa)10.51.0–22.91.0–60.5207
Sediment water content (% mass)13719–4285–586205
Sediment organic matter content (% mass)16.72.0–58.30.5–72.0201
Sediment salinity18.26.6–25.73.3–28.7205
Sediment inline image concentration (mg g−1)0.480.20–1.180.18–1.50205
Sediment inline image concentration (mg g−1)1.190.13–5.160–12.2205

Redox potential values were low (c.−200 mV) at the lowest elevations and increased strikingly with elevation up to c. +400 mV at c. +2.7 m, but did not increase further above this (Fig. 2a). Notwithstanding this clear relationship, there was considerable variation in redox potential, particularly at intermediate elevations (+2.4 to +2.7 m). Although sediment water content was significantly negatively correlated with redox potential (rs = −0.360, < 0.001), as a result of oxic sediments being mainly those with low (<50%) sediment water content, the overall relationship was weak (Fig. 2b) with considerable variation in redox potential across the whole range of water content.

Figure 2.

 The relationships between sediment redox potential and (a) elevation relative to UK reference mean sea level (ODN); and (b) gravimetric sediment water content at Brancaster managed realignment salt marsh, Norfolk, in 2006 (rs = −0.360, < 0.001). The 207 sample points were arranged on a regular, square grid (with 15-m intervals), covering all of the site.

The environmental variables were intercorrelated, as anticipated. Nitrate and ammonium concentrations were highly correlated (rs = 0.996, = 206, < 0.001), so ammonium was omitted from further analysis. Water content, organic matter (loss on ignition), nitrate concentration, salinity, shear strength and sediment depth were all significantly correlated with elevation (< 0.005). In contrast, the correlation between elevation and phosphate content was not significant.

Halophyte distribution

Seven halophytic species had achieved reasonably high frequency or cover at Brancaster MR 4 years after the restoration of tidal flow, but there was still a considerable area of bare ground (Table 2). The distribution of bare ground and that of the seven commonest individual species in relation to elevation and redox potential show clear trends (Fig. 3). Bare ground decreased from 100% cover at the lowest levels to <10% in quadrats higher than +3 m (Fig. 3o). Salicornia showed a unimodal distribution, being commonest in quadrats located between +2.25 and +2.75 m, and averaged <10% cover in quadrats higher than this (Fig. 3m). Puccinellia, Suaeda and Aster tripolium L. (hereafter Aster) (Fig. 3g,i,k) also showed unimodal distributions, with peak abundance occurring slightly higher on the marsh than for Salicornia (cf. Suaeda +2.5 to +3.0 m; Aster and Puccinellia +2.75 to +3.0 m). Limonium vulgare Mill. (hereafter Limonium), Atriplex and Elytrigia were progressively more abundant with increasing elevation, with Elytrigia virtually absent below +2.75 m (Fig. 3e,c,a).

Table 2.   Frequency (%) and mean (SD) cover (%) of bare ground and the plant species occurring on Brancaster managed realignment salt marsh in 2006 (= 208)
 Frequency (%)Cover (%)
Bare ground9138 (33)
Suaeda maritima (L.) Dumort.7815 (19)
Salicornia europaea L.7723 (27)
Atriplex portulacoides L.286 (17)
Aster tripolium L.231 (1)
Puccinellia maritima (Huds.) Parl.222 (6)
Elytrigia atherica (Link) Kerguélen2112 (28)
Limonium vulgare Mill.151 (3)
Sarcocornia perennis (Mill.) A.J. Scott131 (5)
Phragmites australis (Cav.) ex Steud.121 (8)
Suaeda vera Forssk. ex J.F. Gmel.61 (7)
Atriplex prostrata Boucher ex DC.6<1 (1)
Spartina anglica C.E. Hubb.5<1 (1)
Triglochin maritima L.5<1 (1)
Spergularia marina (L.) Besser5<1 (1)
Juncus gerardii Loisel.51 (8)
Spergularia media (L.) C. Presl3<1 (2)
Juncus maritimus Lam.31 (6)
Glaux maritima L.3<1 (3)
Parapholis strigosa (Dumort.) C.E. Hubb.2<1 (3)
Limonium bellidifolium (Gouan) Dumort.2<1 (0)
Festuca rubra L.1<1 (3)
Figure 3.

 The distribution of species abundance (cover or frequency) and bare ground (cover) at Brancaster managed realignment salt marsh, Norfolk, in 2006: (a, b) Elytrigia atherica, (c, d) Atriplex portulacoides, (e, f) Limonium vulgare, (g, h) Puccinellia maritima, (i, j) Suaeda maritima, (k, l) Aster tripolium, (m, n) Salicornia europaea and (o, p) bare ground, in relation to elevation relative to UK reference mean sea level (ODN) (a, c, e, g, i, k, m, o) and sediment redox potential (b, d, f, h, j, l, n, p). Cover values are means. The 207 sample points were arranged on a regular, square grid (with 15-m intervals), covering all of the site.

Comparison with the equivalent patterns of distribution in different redox potential classes (Fig. 3) reveals similar trends but with some important differences from those shown by elevation. At redox potentials lower than 200 mV, bare ground predominated, but it had a much lower cover in sediments that were more oxidized than this (Fig. 3p). Salicornia was able to tolerate redox potentials lower than −100 mV and covered c. 40% of the area in quadrats with redox values in the range 0–200 mV (Fig. 3n). Suaeda was rare when redox potential was lower than 100 mV (Fig. 3j); Atriplex and Limonium were absent below this threshold (Fig. 3d,f) and Elytrigia was absent below 200 mV (Fig. 3b). Puccinellia was widely distributed and appears to be highly tolerant of reducing conditions. It was more frequent when redox potential was >−100 mV, but when the effect of elevation was removed statistically by HP, its occurrence showed an inverse relationship with redox potential.

Hierarchical Partitioning provides estimates of the independent contribution of each environmental variable to the cover of bare ground and the abundance of five of the commonest individual species (Table 3). Overall generalized linear models for Aster and Limonium were not significant. Models for the other species and bare ground were significant and explained 21.4–66.9% of the variation. The relationship with elevation was the most important contribution influencing bare ground and four of the five common species (Table 3); this relationship was quadratic for Salicornia, Suaeda, Puccinellia and Atriplex, but Elytrigia and bare ground showed linear responses to elevation that were positive and negative, respectively. The relationship with redox potential was the greatest influence on the abundance of Suaeda, as well as being a substantial influence on bare ground and the remaining species, with the important exception of Salicornia. Soil nitrate content proved to be an important negative influence on Puccinellia. Sediment depth made a substantial contribution to the effects on Atriplex and Elytrigia, but salinity had greater negative effects on Puccinellia and Atriplex.

Table 3.   Hierarchical Partitioning of the relationships between environmental variables and cover (%) of bare ground, Salicornia europaea and Suaeda maritima, and frequency of Puccinellia maritima, Atriplex portulacoides and Elytrigia atherica
Dependent variableFraction of regression relationship attributable to independent variables (%) according to Hierarchical PartitioningR2 value for GLM regression (%)
ElevationRedox potentialSediment depthSalinityWater contentNitrateShear strengthLoss on ignitionPhosphate
  1. The overall R2 is derived from a Generalized Linear Model (GLM) using all the variables measured. The explained variation is partitioned into the individual contributions of each of the nine variables by Hierarchical Partitioning (HP) (each contribution is a percentage of the overall R2). Non-significant contributions are omitted. Inverse relationships in a GLM regression are indicated by prefixing the percentage with a minus sign, and parentheses around the values for elevation indicate a peak of species abundance/occurrence within the elevation range sampled (a quadratic relationship).

Bare ground−48.5−15.8−8.98.85.0 −3.3−5.0 41.8
Salicornia europaea(66.5)   5.8 8.45.8 36.8
Suaeda maritima(13.5)28.2−4.5   11.2  35.5
Puccinellia maritima(20.2)−11.3 −16.210.0−18.9   21.4
Atriplex portulacoides(29.3)14.6−13.9−17.4−8.06.2   21.7
Elytrigia atherica53.011.1−10.3−4.9−9.66.0 2.6 66.9

Discussion

Coastal engineering projects can provide large-scale, unintended experiments for the study of salt marsh development (e.g. Castellanos, Figueroa & Davy 1994). Managed coastal realignment in response to rising sea levels offers new insights into the environmental tolerances of salt-marsh species because marshes may be initiated at a range of different elevations. Natural salt marshes are typically initiated on low-lying mud or sand flats and then develop through the accretion of sediment. Colonization by pioneering halophytes is important in promoting the trapping of sediments, and the resulting increase in elevation facilitates the establishment of later-successional species (e.g. Ranwell 1964, 1974; Figueroa et al. 2003). The Brancaster managed realignment site is instructive because it proved to be entirely above the level of mean high water neap tides prior to the restoration of tidal flow. The elevational range was comparable with that of mature, natural marshes in the same area and the site was colonized within 4 years by virtually all of the halophytes in the local species pool (Mossman et al. 2012). However, the establishment of species typical of middle and upper elevations was not dependent on sediment accretion and facilitation by flooding-tolerant pioneer species. Consequently, species distributions were uncoupled from the outcome of the more usual primary successional processes, serving to highlight the environmental tolerances of individual salt marsh species. Although bare ground was predominantly at low elevation, it was associated with a wide range of sediment redox potential, and the continued availability of uncolonized ground in the majority of quadrats meant that the influence of interspecific competition on species distributions was reduced, and so environmental factors were acting directly on species distributions rather than modulating competition.

Even with weak potential for biotic interactions, there was nevertheless good evidence for an ‘elevational niche’ (Gray & Mogg 2001) for the seven most common colonists. Elytrigia was confined to the upper elevations; Atriplex and Limonium declined sharply in frequency at lower elevations; Puccinellia, Suaeda and Aster were predominantly in the mid ranges, and Salicornia was mainly on the lower parts of the marsh. The cover of bare ground was also inversely related to elevation. Zonations of species in relation to tidal inundation are a common feature of natural marshes (e.g. Chapman 1938, 1960; Rozema et al. 1988; Zedler et al. 1999; Gray & Mogg 2001; Bockelmann et al. 2002; Silvestri, Defina & Marani 2005), but discrimination between the effects of elevation per se and the environmental factors correlated with it has proved difficult. Tolerance of reducing sediments has been recognized as one of the key determinants of lower elevational limits, as the more frequent and prolonged flooding associated with low elevation in the tidal range increases the likelihood of sediment waterlogging and anoxia (e.g. Armstrong et al. 1985; Castellanos, Figueroa & Davy 1994). Anoxia presents a range of challenges to plants, particularly in saline environments (Colmer & Flowers 2008).

The correlation between elevation and redox potential at Brancaster MR was not perfect; although the lowest redox potentials were found at low elevation and the highest elevations were consistently oxic, there was considerable variation in redox potential at intermediate elevations. There was a surprisingly weak association between sediment redox potential and water content, possibly because of variations in texture and organic content of the sediments. However, it is likely that texture and marsh microtopography (Varty & Zedler 2008) would affect the frequency and duration of water retention, thus explaining the partial uncoupling of elevation and sediment redox potential. Such uncoupling at this site was the key to discriminating their independent effects by Hierarchical Partitioning (Chevan & Sutherland 1991).

As expected, elevation and redox potential were the strongest influences on most species for which a significant model was obtained at Brancaster MR. However, there were important differences between responses to the two variables. The highly significant independent effect of elevation on bare ground may reflect greater mechanical disturbance by tidal scour at lower elevation (Brereton 1971; Tessier, Gloaguen & Lefeuvre 2000) and reduced opportunities for photosynthesis with prolonged inundation (Chapman 1960; Hubbard 1969). Salicornia, an annual halophyte that is regarded as a rapid primary colonist of low marshes in many parts of the northern hemisphere (Davy, Bishop & Costa 2001), was found over the full range of redox potential, which therefore had no independent effect. Its remarkable tolerance of the concentrations of sulphide in the root environment generated at low redox potential (Ingold & Havill 1984; Pearson & Havill 1988) is probably an important factor in regulating its distribution. Puccinellia also tolerated a wide range of redox potential but, because it tended to occupy the more reduced sites at higher elevation, the independent effect of redox potential was negative (i.e. its presence was greater at low redox potential). It is typically a low-marsh species (Gray & Scott 1977; Langlois, Bonis & Bouzillé 2003) whose growth and competitive ability have been stimulated in experiments by flooding (Gray & Scott 1977; Cooper 1982; Huckle, Potter & Marrs 2000). Its tolerance of anoxic conditions is likely to be related to production of aerenchymatous adventitious roots (Justin & Armstrong 1983), notwithstanding a reported sensitivity to sulphide (Ingold & Havill 1984). It is not clear why it should have been independently associated with low concentrations of nitrate in the sediment.

In contrast, the abundance of Suaeda was low at reduced redox potential independently of the contribution of elevation, even though it is a colonizing species frequently found at low elevation. Sensitivity to strongly reducing conditions is consistent with reports of its poor survival in both natural and artificial depressions that experience periodic flooding and impeded drainage (Tessier, Gloaguen & Lefeuvre 2000; Tessier, Gloaguen & Bouchard 2002). Suaeda was also independently influenced by sediment shear-strength, with low abundance on softer sediments, which accords with the tendency of its seedlings to be washed away by tidal action (Tessier, Gloaguen & Lefeuvre 2000). Atriplex appeared to be excluded from sediments of low redox potential, independently of a distribution peak at the mid-elevations. This is in agreement with previous experimental work on this shrubby potential dominant suggesting intolerance of reducing conditions (Mohamed 1998; Crooks et al. 2002) and its common association with well-aerated creek banks in mature marshes (Chapman 1950). Armstrong et al. (1985) reported that it lacked the structural adaptations for flooded soils. Elytrigia is essentially an upper-marsh species (Rozema et al. 1988) that is known to occupy well-aerated sites (Armstrong et al. 1985). Bockelmann & Neuhaus (1999) found that its lower elevational limit was constrained by competition from Atriplex, but our findings indicate that this interaction modifies an ultimate limit set by intolerance of reducing conditions, with Atriplex being marginally more tolerant than Elytrigia.

Tolerance of salinity has been assumed widely to be the main driver of salt marsh zonation. For instance, Rozema et al. (1988) ranked an experimentally derived index of salt tolerance (incorporating flooded and freely drained treatments) for 16 species and obtained a significant correlation with the rank order of zonation in the field. Although HP revealed significant independent negative influences of sediment salinity on Puccinellia, Atriplex and Elytrigia, and a positive influence on bare ground, overall, salinity explained less variation than either elevation or redox potential. Far from there being any facilitation associated with accretion, sediment accretion subsequent to tidal re-activation had an adverse independent effect on the abundance of Suaeda, Atriplex and Elytrigia. This may represent seedling mortality caused by burial with accreting sediment, previously identified for Suaeda by Tessier, Gloaguen & Lefeuvre (2000).

The explanation of species’ distributions on environmental gradients is fraught with difficulty because of the inter-correlation of environmental variables. Yet, small-scale experiments to define the environmental tolerances of individual species fail to capture the complexity of the field. A large-scale field manipulation, combined with HP, has helped to disentangle some of the factors underlying salt-marsh zonation. Although elevation in the tidal frame is undoubtedly the major ultimate determinant of zonation, discrimination between the abiotic factors correlated with it has clarified both the mechanisms of limitation and the fundamental niches of individual species. Even though Salicornia is characteristic of the lowest elevations, it is not restricted by redox potential. The considerable tolerance of reducing conditions by Puccinellia allows it to thrive in e.g. poorly drained depressions at high elevation. The exclusion of Suaeda and Atriplex from reducing sediments explains their exceptional abundance at relatively low elevations in unusually well-drained situations, and limitation of Elytrigia to oxidizing conditions also corresponds with its landward distribution. Thus, quantitative field evidence shows that interpretations of plant distribution based on solely elevational range represent an over-simplification of the environmental heterogeneity of salt marshes. The methodology employed here allows our conclusions to be based on explicit and repeatable assessment of field data and therefore is likely to be applicable to the investigation of species’ distributions on other environmental gradients.

Acknowledgements

We thank the UK Environment Agency (Anglian Region) for allowing access to the study site. We are grateful to NERC for studentships to M.J.H.B. and H.L.M., and to Toby Gardner for helpful discussions about Hierarchical Partitioning.

Ancillary