Present address: Centre for Ecology & Hydrology; formerly the Institute of Terrestrial Ecology, Banchory Research Station, Hill of Brathens, Glassel, Banchory AB31 4BY, Kincardineshire, Scotland (e-mail firstname.lastname@example.org)
Laboratory of Plant Ecology, University of Groningen, PO Box 14, 9750 AA Haren, the Netherlands; and
Institute for Systematics and Population Biology, University of Amsterdam, Kruislaan 320, 1098 SM Amsterdam, the Netherlands
1 The effects of resource competition and herbivory on a target species, Triglochin maritima, were studied along a productivity gradient of vegetation biomass in a temperate salt marsh.
2 Transplants were used to measure the impact of grazing, competition and soil fertility over two growing seasons. Three parts of the marsh were selected to represent different successional stages; Triglochin reached local dominance at intermediate biomass of salt-marsh vegetation. At each stage, three competition treatments (full plant competition, root competition only, and no competition) and three grazing treatments (full grazing, no grazing on Triglochin, and no grazing on Triglochin or neighbours) were applied to both seedlings and mature plants.
3 Competition and herbivory reduced biomass and flowering of Triglochin. The impact of grazing was strongest at the stage with the lowest biomass, while both herbivory and competition had a significant impact at the stage with the highest biomass. When plants were protected from direct herbivory, competition operated at all three successional stages.
4 Grazing reduced light competition when vegetation biomass was low or intermediate, but at high biomass there was competition for light even when grazing occurred. Herbivore exclusion increased the effects of plant competition. Except at low biomass, the negative impact of plant competition on Triglochin performance was greater than the positive effect of not being grazed.
5 Grazing played a minor role in seedling survival and establishment which were largely controlled by competitive and facilitative effects.
6 Once established, the persistence of Triglochin will be determined largely by grazing. Intense grazing in the younger marsh and increasing competition for light in the older marsh will restrict the distribution to sites with intermediate biomass.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In their discussion of the roles of resource competition, herbivory, and predation in relation to ecosystem productivity, Hairston et al. (1960) hypothesized that plant competition will be relatively important at both low and high productivity because herbivore impact is low under these conditions. Plant ecologists, however, have often assumed a constant level of herbivory across productivity gradients, and have arrived at different predictions regarding the role of plant competition. Grime (1979) predicts that competition increases with productivity, while others predict more or less constant competition along a productivity gradient, but changing from competition for nutrients at low productivity to competition for light at high productivity (Tilman 1988; Huisman & Weissing 1995). Experimental tests of these hypotheses have been inconclusive (Gurevitch et al. 1992; Goldberg & Barton 1992; Grace 1993; Goldberg & Novoplansky 1997).
Recently, theoretical models have been developed that include both plant competition and herbivory (e.g. Grover 1995; Leibold 1996; Huisman et al. 1999). Those that consider interactions between the effects of both herbivory and competition for nutrients on plant growth, predict that grazing pressure increases with increasing productivity (Grover 1995; Leibold 1996). This is accompanied by a shift in community composition from plant species that are good competitors for nutrients to those that can tolerate or avoid grazing. Huisman et al. (1999), on the other hand, consider the interactions between herbivory and competition for light and predict that, at high productivity, the smaller plant species with higher forage quality that are preferred by herbivores will be outshaded by taller unpalatable plants. Competition for light should then be most evident where high productivity and a reduction in grazing pressure both contribute to increased biomass.
We have been able to find only one study where the effects of both competition and herbivory on plant performance have been studied across a productivity gradient (Bonser & Reader 1995). However, the findings are limited because young transplants (4–5 week old) were used. Differences in plant stature relative to those of neighbouring plants can lead to different competitive outcomes (Goldberg 1990) because relatively small (e.g. young) plants can only respond to resource levels set by neighbours, while plants of relatively large stature can themselves deplete resources. Moreover, only direct effects of herbivory on the target plant were measured (by protecting individual plants in cages). Although herbivores have a direct effect on plants by removing biomass, they may also cause indirect effects by modifying resource availability (e.g. herbivory may lead to enhanced light penetration through the canopy) or by differential grazing on neighbouring plants.
Our study of the role of resource competition and herbivory on the island of Schiermonnikoog, the Netherlands, was therefore designed to distinguish between direct and indirect effects of herbivory, and to quantify the effects of competition and herbivory on plants at different stages of their life cycle. This island supports a natural productivity gradient created by vegetation succession and, therefore, offers an ideal environment to study interactions between competition and herbivory (Olff et al. 1997). Although there is virtually no predation on the major herbivores (geese, hares and rabbits), grazing pressure does not increase throughout the gradient (as predicted by Hairston et al. 1960) but is highest at intermediate productivity before falling to relatively low levels (Van de Koppel et al. 1996). The decrease at high productivity is due to the abundance of tall, relatively unpalatable plant species, and the accumulation of dead plant material (Van de Koppel et al. 1996; Olff et al. 1997; Van der Wal et al. 1998) and fits the predictions of Huisman et al. (1999).
We conducted a field experiment to test our hypothesis that herbivory has its greatest impact on plant performance at low to intermediate productivity, whereas competition for light should be an important factor controlling plant biomass at high productivity. Triglochin maritima (Juncaginaceae, hereafter called Triglochin), a preferred food plant of herbivores in our study area, was transplanted into three successional stages along the productivity gradient. The impact of competition and herbivory was determined for both seedlings and mature plants of this target species. To discriminate between direct and indirect herbivore effects, cages were used to protect only the individual Triglochin plants against herbivores, whereas exclosures were built to protect both Triglochin and its neighbours.
Materials and methods
Gradual extension of the salt marsh at Schiermonnikoog, the Netherlands, (53° 29′ N, 6° 18′ E) due to natural sedimentation processes has created a gradient encompassing different stages of vegetation development (Olff et al. 1997). The middle–high marsh is characterized by a mosaic of plant communities following the geomorphological pattern of small rises in elevation (5–25 cm) that are separated by shallow gullies. Where Triglochin occurs, it is most abundant on the slopes of these rises. We positioned experimental plots on such slopes at points in the succession that were 15, 30 and 40 years old. Elevation, which determines the frequency and duration of inundation of sites by sea water, was similar for all three successional stages (Table 1). In the two younger stages, vegetation cover was low and the amount of bare soil substantial, whereas in the oldest stage the vegetation was bushy and dense, with little bare soil present (Table 1). The changes in vegetation composition at these three stages represent the successional pathway for the whole middle–high salt marsh (Olff et al. 1997) and closely resemble the pattern observed in other temperate salt-marsh systems (Van Wijnen & Bakker 1997).
Table 1. Description of the three successional stages. Means and standard deviations are given in brackets. Nitrogen mineralization measurements taken from Van Wijnen et al. (1999)
Estimated age (years)
Clay-layer thickness (cm)
Elevation (m + MHT)
N-mineralization rate (g m−2)
Bare soil (% cover)
Dominant plant species
Co-dominant plant species
Triglochin maritima, a rhizomatous rush-like herb, is found almost exclusively on saline or brackish substrates (Davy & Bishop 1991). On Schiermonnikoog, it is found mainly at mid-successional stages (Fig. 1a). It is a perennial species, which forms circular clonal clumps with a high shoot density. It produces long flowering scapes; inflorescences are racemes consisting of many flowers producing numerous seeds. Leaves are thin and erect, and up to 30 cm long. Numerous roots grow from horizontal rhizomes. Leaves and inflorescences are heavily grazed by brent geese (Branta bernicla), barnacle geese (Branta leucopsis), brown hares (Lepus europaeus) and rabbits (Oryctolagus cuniculus). Although goose grazing is the most intense, it is largely restricted to April and May, whereas lagomorphs graze throughout the growing season (from April to October) as well as during winter. At the end of the growing season, above-ground parts of Triglochin senesce and die.
Grazing pressure for each successional stage was determined by calculating the annual food consumption of all herbivores and relating it to live maximum plant standing crop (Van de Koppel 1997). Dropping counts, made between November 1993 and November 1994 (Van de Koppel et al. 1996), were used to estimate consumption. Total grazing pressure (expressed as the consumed percentage of maximum standing crop) was highest in the youngest stage, and dropped to a low level in the oldest stage (Fig. 1b). Day-long counts from observation towers showed a similar pattern only for brent geese (about 16 000, 11 000 and 6000 goose hours km−2 day−1 in the 15, 30 and 40-year-old stages, respectively, in May 1995 (Van der Wal et al. 2000a).
The resources: nitrogen and light
Most temperate salt marshes are nitrogen-limited (Jefferies & Perkins 1977; Kiehl et al. 1997; Van Wijnen & Bakker 1999). In the course of vegetation succession, nitrogen accumulates in the system (Olff et al. 1997; Van Wijnen & Bakker 1997) and above-ground plant standing crop increases, indicating that the successional sequence can be considered as a natural productivity gradient (Van de Koppel et al. 1996). This paper describes the early phase of the productivity gradient in which the nitrogen content of the top soil (which is strongly correlated with the thickness of the top soil clay layer; Olff et al. 1997; Van Wijnen & Bakker 1997), the nitrogen mineralization rate, and the natural (i.e. unfertilized) above-ground plant biomass all increase significantly (Table 1, Fig. 1c). Plant biomass was found to be an appropriate predictor of plant productivity in a salt marsh on the neighbouring island of Terschelling, where net above-ground primary productivity was linearly related to above-ground plant biomass, as measured in 14 plant communities, including those represented in our study (F1,13 = 233.1, P < 0.0001, R2 = 0.95, derived from Kettner 1972).
Above-ground biomass was measured at the end of the experimental period (September 1996) in both fertilized and unfertilized areas of ungrazed vegetation. An area of 10 × 30 cm was clipped to the ground with electric shears and the plant material was dried at 70 °C and weighed. Nitrogen appears to limit plant growth at our youngest two stages, since fertilization (see later for application methods) significantly enhanced above-ground plant biomass (Fig. 1c; paired t-test: Stage 1: t = − 3.25, n = 10, P < 0.01; Stage 2: t = − 4.44, n = 10, P < 0.001), but no effect could be detected in the oldest stage (paired t-test: t = − 0.08, n = 10, n.s.).
In mid-June 1995, the amount of light penetrating the canopy decreased with increasing marsh age (Fig. 1d). About 95% of the incoming photosynthetically active radiation (PAR in moles m−2 s−1) reached the soil surface in the youngest stage, compared with only 45% in the oldest stage. A 2 × 2 × 100 cm long aluminium rod that enclosed 10 PAR-photocells was used to measure light levels in the vegetation.
An overview of the experimental design is given in Fig. 2. Individual seedlings or mature Triglochin plants were transplanted into experimental plots at all three successional stages. Above- and below-ground effects of neighbours were measured according to the experimental design of Wilson & Tilman (1991) by either growing transplants in the presence of all neighbours, without any neighbours, or with roots but not shoots of neighbours. To test for soil nitrogen limitation, each unit consisted of replicate fertilized and unfertilized areas. A solution of NH4NO3 (or equivalent amount of H2O as a control) was applied in early May and June (equivalent to 2 × 10 g = 20 g N−2year−1 available to both Triglochin and neighbours) in 1995 and 1996. One of three grazing treatments was superimposed on each set of fertilization and competition treatments. Either grazing on Triglochin and neighbours was excluded (exclosure treatment), or only Triglochin was protected (cage treatment), or the entire vegetation was accessible to grazing (control). Each mature plant was transplanted to the centre of a 50 × 50 cm plot and grown for two seasons. One seedling was transplanted into each 25 × 25 cm plot at the beginning of each of two growing seasons. To allow subsequent re-identification, a very thin 1.5 cm diameter plastic ring was placed around transplanted seedlings. All treatments were replicated 10 times. In total, 540 tussocks of mature plants and about 1000 seedlings were transplanted.
Given the scale of the experiment and the limitations posed by the available habitat on the island, we were not able to replicate successional stages. This experiment thus suffers from pseudo-replication at the level of successional stage/productivity and the findings should be viewed accordingly. However, the 10 blocks of replicates within each successional stage were spread out over as large an area as possible (3, 7 and 6 ha for the 15, 30 and 40-year-old stage, respectively) to allow for spatial variation. The three successional stages were 1.5 km apart.
Seeds were collected in 1994 and 1995 from plants of Triglochin in the 30-year-old marsh. One-year-old seeds were germinated in a glasshouse and grown for 3 weeks in a mixture of half sand and half peat and transplanted into the field without attached potting soil. The first transplants (one seedling per plot) were placed in position at the end of April 1995, at which time germination in the field had already taken place (Hutchings & Russell 1989). All transplanted seedlings died due to prevailing dry, warm weather and were replaced in the last week of May. These were then harvested in July 1995 after a growing period of 2 months. We repeated the experiment in early May 1996 and harvested in September 1996 after a growing period of 4 months.
In early April 1995, mature Triglochin plants in the 30-year-old marsh were excavated with a soil corer (7 cm diameter) to a depth of approximately 10 cm, and transplanted (with the cylinder of soil and roots intact) directly into all three successional stages. Some minor root damage occurred during excavation. Before transplantation, other plant species were removed from the core by hand. Plants were harvested after two growing seasons in September 1996.
The no-neighbour treatment (NN) was accomplished by removing all above-ground plant parts from the plot with electric shears (50 × 50 cm for mature plants and 25 × 25 cm for seedlings). Roots were not removed in order to prevent severe soil disturbance, as this could lead to adverse effects on plant growth due to salt stress (Srivastava & Jefferies 1995), but regrowth of neighbours was clipped every 2–4 weeks as appropriate. To prevent soil from drying out, green plastic mats were pinned to the ground with metal pins so that the bare soil beneath remained moist throughout the experiment. Transplants with only rooting neighbours (RN) had the shoots of neighbouring plants tied back with nets of black cloth (2 mm mesh). These transplants received 99% of the incident radiation, as measured in June 1996. Nets were fixed on the ground in the centre of a plot with short metal pins, while at all four corners long pins held the net approximately 10–30 cm above the ground, depending on the height of the vegetation. The all-neighbour treatment (AN) consisted of a transplant grown in the centre of a plot, without further manipulation. The competition treatments were not confounded by differences in herbivore grazing pressure.
The proportion of plots in which Triglochin was grazed did not differ significantly among the three competition treatments (NN: 0.49, RN: 0.43, AN: 0.41;
= 2.06, P > 0.3, mean of six inspections performed every other week during the first year of the experiment). The most important species that were associated with Triglochin are listed in Table 1, with plants of low stature in the two younger successional stages and bushy, tall plants such as Atriplex portulacoides and Artemisia maritima (≥ 50 cm height) in the oldest stage.
Herbivores were excluded from plots by erecting exclosures (1 × 2 m) made of chicken wire (mesh 2.5–5 cm). Wire cages were used to prevent grazing on Triglochin by surrounding the tussocks of mature plants (1.5 cm mesh size, 25 cm height) or seedlings (1 cm mesh size, 5 cm height). The cages around mature plants did not completely prevent herbivory so that at high grazing pressure some leaves and especially flower heads of Triglochin were eaten. The control treatment consisted of plots that were unmodified.
At harvest, shoots of adult plants were cut at ground level and roots were then sampled with the soil corer used for transplantation. In the laboratory, all flowering scapes were separated from mature shoots and counted; the few dead leaves were discarded. The root samples were washed and roots of other plant species and the few dead Triglochin roots were removed. In both years, seedlings were collected with roots attached and, after washing off the soil, roots and shoots were separated. All plant material was dried at 70 °C and weighed.
Seedling survival was estimated by counting the number of dead seedlings in July 1995 and in August 1996. Missing seedlings were considered dead when no seedling was present and only the small plastic ring remained. Mature plants were recorded as dead when no live above- or below-ground parts were found. In July 1995, three randomly chosen shoots were collected from each mature tussock. The number of leaves and the length of the longest leaf (to the nearest mm) were recorded for each transplant, and leaves were weighed after drying at 70 °C.
Biomass data for adults and seedlings were analysed using analysis of variance with competition, grazing, fertilization, and successional stage as main factors. The experimental blocks (see Fig. 2) were nested within stage, but no further interactions incorporating ‘block’ were included (Newman et al. 1997). Due to mortality, data on seedling mass were highly unbalanced. All non-significant effects were therefore removed from the analysis step by step, to end up with a model of significant factors only (Crawley 1993), although block effects nested within stage were retained, regardless of their significance. All data were log-transformed prior to statistical analysis to improve homogeneity of variances. We present total sums of squares as an indication of the importance of the different elements in the analysis (Underwood & Petraitis 1993). In some cases (indicated in text), tests were run on subsets of the data to answer specific questions. Data for Triglochin flower heads were Poisson-distributed and Poisson-regression was therefore applied to these data with competition, grazing, fertilization, and stage as factors (Crawley 1993). Analysis of covariance was used to test for differences among stages in the relation between plant height and leaf mass. Seedling survival was tested with logistic regression for 1995 and 1996 separately. As a measure of variance, standard error of the mean is given unless otherwise stated.
Adult survival was high; only 12 of the 540 plants died during the 2 years. Five plants died when transplanted to the 15-year-old successional stage, and of these, four were not protected against grazing (the fifth was caged), but survival was not systematically related to competition. In the 30-year-old stage, the source of the transplants, the only death was a plant subject to full competition and herbivory. At the 40-year-old stage, all six dead transplants were subject to full competition. Four of them were from inside exclosures, while the other two were not protected from grazing.
Most biomass of adult Triglochin plants was below ground. Above-ground tissue contributed only 12% to total plant biomass (SD = 9.1%, n = 525) on a dry-weight basis.
All main factors, except addition of fertilizer, significantly affected total live Triglochin plant biomass (i.e. successional stage, grazing, and competition; Table 2). Grazing explained the most variation (21%), while competition only contributed another 7%. Both grazing and competition negatively affected plant biomass (Fig. 3). Grazed Triglochin transplants were, on average, smaller than ungrazed transplants (ANcontrol: 3.5 ± 0.26 g dry weight per transplant; ANcaged: 5.9 ± 0.26 g dry weight per transplant). The difference in biomass due to grazing was strongest in the marsh at the 15-year-old successional stage, and less pronounced in the two later stages (stage × grazing: P < 0.001). Although exclusion of direct herbivore attack on Triglochin always increased plant biomass (compare ANcontrol with ANcaged), total exclusion of herbivores from the vegetation caused a reduction of Triglochin biomass in the older two successional stages (compare ANcaged with ANexcl). This is probably due to enhanced negative effects of competition. Only in the 15-year-old stage did total exclusion of herbivores lead to a net increase in Triglochin biomass (ANcontrol vs. ANexcl). Biomass reduction due to competition was slightly stronger when plots were fertilized (comp × fert: P = 0.048), but this interaction explained a very small part of the total sums of squares (Table 2). Triglochin biomass was, on average, lowest in the 30-year-old marsh.
Table 2. Results of the analysis of variance on Triglochin biomass (above- and below-ground) of mature transplants. All main effects are given, but non–significant interactions are not presented
% of total SS
Block within stage
Stage × grazing
Grazing × competition
Competition × fertilization
Negative effects of competition were observed when the vegetation surrounding Triglochin was grazed or ungrazed (Fig. 3: cages and exclosures). However, when plants were not protected from direct herbivore attack (control), competitive release did not lead to overall higher plant biomass, probably because of an increase in the amount of plant tissue removed by herbivores (grazing × comp: P < 0.001). Nevertheless, in the 40-year-old stage, tying back neighbours (RN treatment) led to a significant increase in grazed transplant performance, suggesting competition for light (Fig. 3c: grazed plants; one-way anovaF2,57 = 3.8, P < 0.05).
In the absence of direct herbivory, competition acted most strongly when the surrounding vegetation was ungrazed (Cage vs. Excl.: grazing × comp. F2,349 = 5.85, P < 0.01). When Triglochin was caged, competition switched from mainly below-ground competition in the younger successional stages to competition for light in the oldest stage (Fig. 3: Cages; compare AN with RN and NN for the three stages separately). In totally ungrazed environments (exclosures), however, both above- and below-ground competition played a significant role.
Triglochin produces flowering scapes that form the tallest parts of the plant. Herbivores were able to graze down flower heads even when individual plants were caged. Therefore, a statistical comparison was made between grazed and exclosed plants only, excluding caged plants from the analysis.
All four main effects explained significant amounts of the overall variance, and interacted significantly with each other (Poisson-regression: stage × grazing × comp × fert: χ2 = 14.8, P < 0.01; Fig. 4). To gain insight in the most important factors, we analysed the flowering data using stepwise forward Poisson-regression. Grazing (23% of total deviance explained) and competition (19% explained) had far stronger effects on the number of flowering scapes than did the successional stage (8% explained) or fertilization (1% explained). Interactions, although significant, did not contribute much to the overall model (stage × grazing: 5%; all others < 2% of total deviance explained).
On average, grazing reduced the number of flower heads from 3.2 ± 0.4 to 0.5 ± 0.1 per transplant (Fig. 4: Control vs. Excl.). Overall, the effect of competition was the same order of magnitude, with 3.6 ± 0.5 flower heads in the absence of neighbours (NN treatment), 1.5 ± 0.2 with rooting neighbours only (RN treatment), and 0.5 ± 0.1 in the presence of full competition (AN treatment). The effects of competition were dependent on grazing treatment. When plants were grazed, the number of flower heads was low both in the absence of neighbours (0.7 ± 0.2), and in full competition (0.1 ± 0.1). In the absence of grazing, however, competition caused a strong reduction in the number of flower heads (6.5 ± 0.8 vs. 0.9 ± 0.2). Furthermore, effects of grazing, competition, and fertilization differed among stages. Poisson-analysis showed that in the 15-year-old successional stage, most of the variance was explained by grazing (Table 3). Grazing and competition were both important in the 30-year-old stage, while competition was most important in the 40-year-old stage. Fertilization had a negative effect on the number of flower heads, but explained only a marginal proportion of the total variance. In the absence of neighbours, grazing reduced the mean number of flower heads by 99%, 71% and 68% for the 15-, 30- and 40-year-old successional stages, respectively. Note that in the youngest stage not a single flower head was found when grazed plants were in full competition with other plants. On average, number of flower heads was lowest in the 30-year-old stage.
Table 3. The proportion of variance (%) explained in the number of flowering heads per transplant for each successional stage. Stepwise forward Poisson analysis was performed for each successional stage, incorporating only factors that were found significant in stepwise backward analysis
Grazing × competition
Grazing × fertilization
Competition × fertilization
Grazing × competition × fertilization
In the 40-year-old stage, plants produced long leaves of low mass. With increasing leaf length, corresponding mass gain was significantly less in the 40-year-old stage than in the 15 and 30-year-old stages (ancova; leaf length × stage: F2,272 = 10.8, P < 0.001). Also, the percentage of the total above-ground plant biomass was significantly higher in the 40-year-old successional stage, compared to the younger two stages (one-way anova: F2,522 = 7.8, P < 0.001). The relative contribution increased from 11.5% ± 0.8 in the 15-years-old stage to 13.6% ± 0.7 in the 40-years-old stage. Both responses suggest strongly that light becomes a limiting resource in the older successional stages.
In seedlings, a similar allocation pattern was found with relatively low investment above ground in the younger successional stages relative to the oldest stage (one-way anova: F2,218 = 23.3, P < 0.0001). The percentage of the total above-ground biomass increased from 40.9% ± 1.7 and 42.3% ± 1.6 in the younger two stages to 53.5% ± 1.4 in the 40-year-old stage.
Survival of Triglochin seedlings was comparable between the 2 years: 42% in 1995 and 49% in 1996 (Table 4). Significant effects on survival were found among stages, grazing treatments, and competition treatments (Table 5), although the grazing effect in 1995 varied with stage. Nutrient supply had no effect on seedling survival.
Table 4. Seedling survival within each successional stage as a proportion of initial number in the first (1995) and second (1996) year of the experiment. AN = all neighbours, RN = rooting neigbours, NN = no-neighbours
Table 5. Results of the logistic regression on seedling survival for each of the two years
% of deviance explained
Stage × comp
Stage × grazing
Grazing × comp
In both years, survival was lowest in the youngest successional stage and highest in the oldest stage (Table 4). Differences in survival rate among stages were greatest in 1995, with mean survival averaging 5% in the youngest stage, but 58–62% in the other two stages. The lowest survival was measured in plots unprotected from grazing. In 1995, mean survival was 30% in seedlings exposed to herbivory, but 48% and 46%, respectively, when protected by cage or exclosure. However, survival in the youngest successional stage was low, irrespective of the grazing treatment (1995: stage × grazing: P < 0.0001). Mean survival in 1996 was 37% of seedlings in grazed plots and 55% of seedlings in both caged and exclosed plots.
The effects of neighbours on seedling survival were not straightforward. Seedling survival was highest when above-ground parts of neighbours were tied back (RN 11% and 27% higher than AN in 1995 and 1996, respectively), although in 1995 this effect was obscured because of very low overall survival in the youngest stage (stage × comp. P < 0.01). The lower survival rate in the absence of neighbours was not due to higher grazing pressure as these low survival rates also occurred in cages (1996: NNcontrol vs. NNcaged). However, exclusion of herbivory slightly improved seedling survival when seedlings were in full competition (ANcontrol vs. ANcaged), although not at the level that was reached with rooting neighbours only. In the absence of herbivory, the opposite trend was observed (grazing × comp. P < 0.01). Full competition reduced survival to levels far lower than in the absence of neighbours in 1996, but not in 1995. Apparently, the presence of neighbours induced competition for light (AN vs. RN), whereas the total absence of neighbours may have enhanced negative effects of abiotic conditions (NN vs. RN), such as high energy waves and salt stress.
Competition and successional stage were significant factors in both 1995 (comp F2,194 = 15.4, P < 0.001, 10% of total SS explained; stage F1,194 = 15.2, P < 0.001, 11% explained) and 1996 (comp F2,193 = 19.7, P < 0.001, 15% explained; stage F1,193 = 3.3, P < 0.05, 2% explained). Plants were, on average, biggest in the 15-year-old stage, although the differences were small (Fig. 5). In 1995, this observation was based on only a few surviving seedlings. In 1996, however, a similar pattern was found, and was now based on a far larger sample size. Seedling mass in 1995 was 1.71 ± 1.21 mg in the 30-year-old successional stage but only 1.02 ± 0.67 mg in the 40-year-old stage, whereas no differences between these two stages were found in 1996.
In 1995, grazing influenced seedling biomass (F2,194 = 6.2, P < 0.01), although the contribution to the overall explained variance was small (4%). Seedlings growing in a grazed environment did suffer from direct grazing effects; seedling biomass was 1.61 ± 0.13 mg when growing in cages and 1.27 ± 0.13 mg when not protected. However, seedling biomass in exclosures was as low as in control plots (1.27 ± 0.10 mg).
Neighbours had a negative impact on seedling performance in both years, causing approximately 50% reduction in biomass. The negative impact of neighbours was strongest in the youngest stage in 1995. Hardly any effects on seedling biomass were observed between plants without neighbours and plants growing with rooting neighbours only (Fig. 5), suggesting that seedling biomass was mainly reduced by above-ground competition.
We investigated the performance of Triglochin maritima at three successional stages along a gradient in which mineralization rate, soil fertility and plant standing crop increase markedly, so that it can be regarded as a productivity gradient (see also Van de Koppel et al. 1996). The combined effects of competition and herbivory along productivity gradients are poorly studied. As far as we know, this is the first experimental study incorporating the effects of grazing and competition on plant biomass and flowering, and that takes both direct and indirect effects of grazing into account. In contrast to most other studies, we analysed total live plant biomass, instead of only above-ground biomass which, when feasible, is clearly a better reflection of plant performance. Although herbivores consumed only above-ground tissue, below-ground biomass was also strongly affected due to rapid regrowth of Triglochin after grazing (D. Dähnhardt, unpublished results).
Triglochin performance over the productivity gradient
Performance of mature Triglochin plants in the absence of both herbivory and competition was lowest in the 30-year-old marsh, from which all plants were originally collected. As predicted, transplants placed in the more productive older marsh reached a higher final biomass after 2 years. Contrary to prediction, however, transplants grown in the apparently less productive, younger marsh also performed better than those placed back in the 30-year-old marsh. This might be due to site-specific differences, such as soil composition or soil-water availability, but it might also be attributed to the effects of transplantation. Mature plants were transplanted with an attached root ball of sandy clay soil of approximately 10-cm depth, which is more than twice as much as the average clay layer thickness of soil from the youngest successional stage.
Seedlings vs. mature plants
In most studies of the importance of competition or herbivory in plant communities, seedlings have been used rather than mature, established plants. It is argued that seedlings are most susceptible to effects of both herbivory (Reader 1992) and competition (Goldberg 1990). In our study, the conclusions drawn differed between mature plants and seedlings. Competitive and facilitative effects were prominent in the seedling data, whereas mature plants were largely controlled by grazing, with competitive effects manifest only at later successional stages where vegetation biomass is higher.
Seed production and seedling survival
Grazing significantly reduced the number of Triglochin flower heads. This is in agreement with previous work in a salt marsh in eastern England where rabbits removed up to 95% of the flower heads of Triglochin (Davy & Bishop 1991). Similarly, Mulder & Ruess (1998) report from southwestern Alaska on a decrease in sexual reproduction in Triglochin palustris due to geese grazing.
Mean seedling survival increased over the productivity gradient. Survival was highest when seedlings grew in the shelter of neighbours but did not face the negative effects of shading (i.e. rooting neighbours). Furthermore, seedlings in cages had higher survival in the all-neighbour treatment than in the treatment without neighbours, probably indicating facilitation. In contrast, in the less exposed ungrazed vegetation (exclosures), survival was higher for seedlings without neighbours. For those seedlings that survived, however, neighbours had mostly a negative effect on seedling mass instead of a positive, facilitative effect.
Although seedlings may escape from herbivory because of their small size, we found that survival was reduced by grazing. Subsequent seedling performance was primarily influenced by competition with neighbours. Experimental treatments (i.e. grazing, competition, and fertilization) explained much lower proportions of the variance of seedlings than of mature Triglochin plants, indicating that factors other than those subject to experimental control were determining seedling survival and establishment. Most of the discussion will therefore be devoted to the responses of mature plants.
The intertwined effects of competition and herbivory
Herbivory is often seen as a confounding factor when measuring competition among plants (Reader 1992). Experimental removal of neighbours to exclude competition may simultaneously reduce food and shelter available to herbivores. Consequently, reduced herbivory rather than reduced competition may be responsible for the improved performance of a target plant. In contrast, the effects of reduced competition may remain undetected due to an increased rate of herbivory on a target plant after removal of its neighbours. Competitive interactions among plants may also change when herbivory is excluded from both neighbours and the target plant, and is no longer representative of the system under study.
We found no evidence for a reduced rate of herbivory following neighbour removal in mature Triglochin plants. Since grazing pressure on Triglochin was high, one could argue that neighbour removal might enhance the rate of herbivory as the target plants are then easier to find. Competitive release in a grazed system did not lead to a higher biomass of mature plants. However, plants facing similar environmental conditions but growing in cages clearly did respond to competitive release. Grazing pressure presumably increased following experimental neighbour removal, counterbalancing the improved plant performance in the absence of competition. It was only at sites of high productivity, where grazing pressure was relaxed, that effects of competition were actually measurable in unprotected target plants and not totally masked by increased herbivore off-take.
The level of herbivore exploitation was lowest at the highest productive stage, compared to the two stages of lower productivity. Consequently, mature Triglochin plants were grazed less intensively at high productivity. Reduced losses due to lowered rate of herbivory were, however, counterbalanced by a reduction in performance due to competition for light, illustrating the tight interaction between herbivory and competition. At low productivity some mature plants even died when not protected from grazing, whereas at high productivity some plants died while competing for resources with neighbours.
Effects of resource competition
Two bodies of theory predict the role of resource competition in ecosystems. Grime (1979), amongst others, predicts that the role of competition becomes more important with increasing productivity, whereas Tilman (1988) and colleagues expect competition to be equally strong at sites with low and high productivity. We measured the effects of competition on mature Triglochin plants at three stages of a successional gradient in which mineralization rate, soil fertility and plant standing crop increase markedly. We did not detect a marked increase in the intensity of competition over this gradient, but differences were found between caged and exclosed plants. In exclosures, mature Triglochin biomass increased by nearly 50% following removal of neighbours; this was only a 26% increase when caged. Competition intensity ([NN − AN]/NN: Belcher et al. 1995) faced by caged plants tended to be highest at high productivity (0.25, 0.22 and 0.32 in successional stages 1 to 3, respectively), but high variation within experimental blocks prevented appropriate testing. In exclosures, however, mean competition intensity did not increase with increasing productivity (0.46, 0.48 and 0.48 in stages 1–3, respectively). Obviously, Triglochin experienced a lower level of competition when its neighbours were grazed, relative to ungrazed neighbours.
Nature of resource competition
Tilman (1988) and colleagues (Slobodkin et al. 1967; Newman 1973; Grubb 1985) predict that, although competition intensity may be constant, a qualitative shift is expected from competition for soil resources at low productivity to competition for light at high productivity. In our study, mature transplants growing in exclosures appeared to face competition for both soil resources and light in all three stages of the productivity gradient. A shift from below-ground competition at lower productivity to light competition at high productivity was, however, prominent in caged plants growing in an otherwise grazed environment. Grazing probably prevented light competition from becoming operational at lower levels of productivity.
Although competition for light was demonstrated in the absence of grazing on Triglochin (cage and exclosure treatments), experimental bias might occur when grazing of the target plant was excluded. Grazing pressure on plants might differ among the different competition treatments and improved performance of the transplant might then be due to lower herbivore off-take instead of improved light conditions. Our data on grazing did not show differences in attack rate among competition treatments and, therefore, we do not think that improved performance of transplants is an artefact of reduced grazing of the target plant. The phenotypic responses of both seedlings and mature plants appear to confirm our conclusion that light availability is limiting Triglochin growth in the 40-year-old stage.
Direct and indirect effects of herbivory
Although caging increased performance of the target plants, excluding herbivores from both Triglochin and neighbours induced stronger competitive effects, lowering Triglochin performance and indicating that herbivory had indirect effects.
Grazing not only reduces the competitive ability of neighbours, it also reduces vegetation height, leading to enhanced light penetration through the canopy (McNaughton 1992; Huisman et al. 1999). Mulder & Ruess (1998) showed that clipping neighbours resulted in increased biomass of Triglochin palustris. In the case of selective herbivory, however, neighbours might reduce the grazing pressure on palatable plants by making them less obvious, which might ameliorate the direct grazing effects, as was shown in the same study on Triglochin palustris (Mulder & Ruess 1998). Reduction in plant cover can also reduce the potential role for facilitative effects of neighbours, and it may lead to less favourable abiotic conditions for plant growth (Callaway & Walker 1997). Salt marshes are prone to salt stress, water stress, and high energy waves (Adam 1993). Early in succession, where plant cover and biomass are low, abiotic conditions are expected to be most extreme. Here, caged mature Triglochin plants performed worse than plants growing in an ungrazed situation, indicating lack of shelter from neighbours. At the two successional stages where salt marsh vegetation showed higher productivity, however, plants in cages performed better than plants in exclosures, indicating that the competitive effects of neighbours exceeded their facilitative effects.
Under natural conditions, Triglochin abundance was low in both the 15 and 40-year-old successional stages, but reached local dominance in the 30-year-old stage (Fig. 1a). The transplant experiment demonstrated that, in the young marsh, herbivores were able to keep the biomass of mature Triglochin low. Furthermore, seedling survival was reduced by herbivory, as was seed stock. We expect herbivores to be able to prevent expansion of Triglochin in the young marsh and to be the key factor explaining the low abundance of this plant species in the younger marsh. In the older marsh, competition for light limited Triglochin. Although few plants died, we suspect that the 2-year experimental period was too short to lead to a general die-off among the transplants. In the older marsh, Triglochin is largely overgrown by Atriplex portulacoides. Interestingly, Triglochin abundance increases again in much older parts of the salt marsh where cattle prevent competition for light by grazing upon tall plants (Olff et al. 1997).
Long-term grazing effects
The experiments lasted for only two growing seasons. Eventually, exclusion of herbivores will lead to plant species replacement (Drent & Van der Wal 1999; Van der Wal et al. 2000b). Tall-growing plants like Atriplex portulacoides are likely to become dominant in younger stages of the salt marsh as well, and these will then induce competition for light. Despite strong negative effects of herbivores on Triglochin biomass and flowering, in the long run grazing will benefit this plant species by postponing competition for light.
More experimental studies are needed before generalizations can be made about the role of herbivory and competition along gradients of primary productivity (Louda et al. 1990; Goldberg & Barton 1992). This study shows that both grazing and competition play important roles in determining the performance and distribution of plant species, and that these two processes are closely linked: when grazing pressure is relaxed, competition with neighbours is intensified. Grazing is shown to influence competitive interactions among plants, acting both directly on the target plant and indirectly via its neighbours. The significance of competition and herbivory is largely dependent on plant stature relative to the neighbouring vegetation. Although establishment of Triglochin starts from seed, the high grazing pressure on plants of larger size will determine its persistence in the sward. The distribution of Triglochin is ‘sandwiched’ between intense grazing in the younger marsh and increasing competition for light in the older marsh.
We are grateful to Harm van Wijnen, Peter Korsten, Linda Zwiggelaar, Julia Stahl, Saskia Aldershof, Anita van der Wal, and especially Cornelia Rothkegel and Bas Kers for their help in the field or during sample processing. We thank Jelte van Andel, Carsten Dormann, Rudi Drent, Lindsay Haddon, Jef Huisman, Bob Jefferies, Johan van de Koppel, Michael McDonald, E.I. Newman and an anonymous referee for their useful comments on earlier versions of the manuscript.
Received 17 February 1999revisionaccepted 14 October 1999