Interactive effects of soil moisture, vegetation canopy, plant litter and seed addition on plant diversity in a wetland community

Authors

  • Shaojun Xiong,

    Corresponding author
    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
      Present address and correspondence: Shaojun Xiong, Landscape Ecology Group, Department of Ecology and Environmental Science, Uminova Science Park, Umeå University, SE-901 87 Umeå, Sweden (tel. +46 90 7869573; fax +46 90 7867860; e-mail xiong@eg.umu.se).
    Search for more papers by this author
  • Mats E. Johansson,

    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
    2. Landscape Ecology Group, Department of Ecology and Environmental Science, Uminova Science Park, Umeå University, SE-901 87 Umeå, Sweden, and
    Search for more papers by this author
  • Francine M. R. Hughes,

    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
    Search for more papers by this author
  • Adrian Hayes,

    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
    Search for more papers by this author
  • Keith S. Richards,

    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
    Search for more papers by this author
  • Christer Nilsson

    1. Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, UK,
    2. Landscape Ecology Group, Department of Ecology and Environmental Science, Uminova Science Park, Umeå University, SE-901 87 Umeå, Sweden, and
    3. Department of Natural and Environmental Sciences, Mid Sweden University, SE-851 70 Sundsvall, Sweden
    Search for more papers by this author

Present address and correspondence: Shaojun Xiong, Landscape Ecology Group, Department of Ecology and Environmental Science, Uminova Science Park, Umeå University, SE-901 87 Umeå, Sweden (tel. +46 90 7869573; fax +46 90 7867860; e-mail xiong@eg.umu.se).

Summary

  • 1We carried out a factorial experiment to examine how groundwater availability (low and high sites with intermediate or rare flooding), vegetation canopy, leaf litter and seed availability interacted to determine the species richness of a productive wet grassland community in Wicken Fen National Nature Reserve, Cambridgeshire, UK. Seeds of 18 species were added to half the plots in each of eight combinations of elevation, canopy and litter, and seedling emergence was observed for two growing seasons.
  • 2Both individual and interactive effects on plant diversity and colonization were determined for all four examined factors. Interactive effects explained 41–63% of the total variation in both species richness and numbers of individuals growing from added seeds.
  • 3Neither elevation nor vegetation canopy had significant individual effects on total species richness, but their interaction was significant. Litter addition limited seedling emergence at the low elevation but favoured it at the high elevation.
  • 4The relative importance of vegetation canopy and plant litter in affecting plant community composition varied with the community parameter considered (species richness or number of seedlings), elevation and stage of vegetation development. In general, plant litter was more important in determining species richness, whereas the vegetation canopy was more important in determining seed germination and seedling emergence. Plant litter was also more important than vegetation canopy at an early stage of vegetation development and at low elevation.
  • 5Seed availability was the most important factor in determining overall species richness in the studied community. The influence of the local seed bank was very limited. Seedling emergence and seedling species richness were generally enhanced by lower elevation and seed addition, but depressed by vegetation and litter addition.
  • 6The complex relationships observed have considerable implications for ecological modelling and ecosystem restoration. Manipulation of one factor may produce unexpected effects on other factors, which may induce a series of consequences for the whole community. Further knowledge on how natural communities are organized and maintained is needed to guide the management of ecosystems.

Introduction

A variety of abiotic and biotic factors have been proposed as determinants of plant species diversity. However, when considering the rates of species gain or loss in the community, most attention has been directed to the roles of disturbance, physical resources, species interactions and propagule availability (Zobel 1992; Eriksson 1993; Ricklefs & Schluter 1993; Ward et al. 1999). The failure to identify any single factor as the major determinant of species diversity on a more local scale suggests that interactions between several factors are often more important. However, manipulative experiments designed to separate the effects of different factors involved in interactions or to combine different individual effects are rare (but see Lenssen et al. 1999; Zobel et al. 2000; Levine 2001). The more factors that are involved in an experiment, the more complex the possible interactions, and the more sophisticated the analytical techniques required to interpret them.

Many land–water interfaces, for example intermediately to rarely flooded parts of riparian zones in the Northern Hemisphere, have a species-rich flora (Nilsson et al. 1989; Naiman et al. 1993; Ward et al. 1999). These areas are also characterized by high productivity and highly dynamic plant communities, because of disturbances caused by flood erosion, and cycling of water-borne organic matter (plant litter) as well as plant propagules (Nilsson et al. 1999). Little is known, however, about how these factors interact to create and maintain high species richness, or whether some factors are more important than others. Species richness in productive communities has been suggested to be locally determined largely by biotic interactions, primarily competition (Grime 1979; Tilman 1987). This has been demonstrated for old fields (Wilson & Tilman 1991), grasslands (Tilman 1988) and shorelines (e.g. Twolan-Strutt & Keddy 1996; Brewer et al. 1997; Lenssen et al. 1999). The importance of plant litter has probably been underestimated (Facelli & Pickett 1991; Foster & Gross 1997, 1998; Xiong & Nilsson 1999), because production has often been measured as above-ground biomass including both live and dead material (e.g. Wheeler & Shaw 1991; Wilson & Tilman 1991; Gough et al. 1994). Although studies have stressed the importance of leaf litter for the structuring of plant communities (van der Valk 1986; Xiong & Nilsson 1997, 1999; Nilsson et al. 1999), separation and comparison of the effects of plant litter and vegetation canopy have rarely been undertaken (but see Foster & Gross 1998).

Water availability is important in affecting plant colonization and distribution in wetlands (e.g. Menges & Waller 1983; Blom & Voesenek 1996; Johansson & Nilsson 2002; Kellogg et al. 2003) and number of plant niches (Silvertown et al. 1999). Even a small difference in soil moisture may result in a significant difference in seed germination and thus in floristic diversity of a wetland community (Keddy & Ellis 1985; Vivian-Smith 1997). On the other hand, water fluctuation may also regulate the effects of plant litter on the vegetation (Xiong & Nilsson 1997). It is possible that the effects of plant litter are weakened where ground water and soil moisture levels are elevated and the rate of litter decomposition is increased (see Xiong & Nilsson 1997 for further references). Consequently, the relative importance of litter and vegetation may vary with soil water availability.

Propagule availability is another key factor for plant colonization and species composition (Pickett & White 1985). Turnbull et al. (2000) noted the importance of seed limitation in ecosystems ranging from sand dunes to forest communities, and Eriksson & Ehrlén (1992) and Zobel et al. (2000) proposed that seed limitation is as important as species interactions in determining species diversity in grasslands. Contrary to upland ecosystems, however, riparian zones often receive large numbers of plant propagules gathered and transported by floods (Nilsson & Grelsson 1990; Andersson et al. 2000), suggesting that the seed species pool may already be saturated. By studying the development of restored wetland communities, Kellogg & Bridgham (2002) found that initial planting and seeding had a small effect, whereas a restoration of the hydrological processes resulted in a more variable plant community. This suggests a more important influence of abiotic factors than of seed bank on plant colonization and species composition (cf. van der Valk 1981). Thus, riparian communities may be limited by habitat conditions rather than by seed availability. It is therefore important to know whether and how seed addition affects an existing plant community, and how vegetation canopy, plant litter and soil water conditions influence the colonization of emerging seedlings.

In this study, we examined experimentally the relative importance of proximity to groundwater, vegetation canopy, leaf litter and seed addition in determining species richness of a productive wet grassland community. We designed and established a factorial experiment that was composed of 16 treatments among 64 plots, each of 0.5 × 0.5 m, in Wicken Fen National Nature Reserve, Cambridgeshire, UK. The emergence from seeds of 18 species (1.5 times the mean number of species per plot) added in each plot was observed for two growing seasons. We hypothesized that: (i) both litter and vegetation canopy are important in determining plant colonization and species richness in riparian zones, but that their relative importance changes along the hydrological gradient; and (ii) species diversity in the studied community is primarily limited by the availability of colonizable sites rather than by external seed input.

Materials and methods

study site and experimental design

Wicken Fen (52°21′ N, 0°15′ E) is a National Nature Reserve that preserves a remnant of formerly extensive topogenous (fen) and ombrogenous (bog) mires in eastern England in an area known as ‘The Fens’. The Fens have been reclaimed since Roman times, first for grazing and then for arable agriculture, but the most intensive period of reclamation began in the 17th century and has continued since then. Wicken Fen has an exceptionally rich fauna and flora and was first set up as a nature reserve in 1899 by the National Trust.

A stretch along a lake shore with little topographic variation was chosen for the field experiment. Local vegetation was dominated by herbaceous species such as Cirsium arvense, Elytrigia repens, Glechoma hederacea, Potentilla anserina, Arrhenatherum elatius and Holcus lanatus. The soil consisted of well-drained peat (as deep as c. 100 cm).

The experiment was set up on 22 April 2001. Four experimental factors, each represented by two levels, were combined in a factorial design: elevation (low vs. high), vegetation canopy (intact vs. removed), litter (added vs. removed) and seed availability (sowing vs. non-sowing). In total, 16 treatments, each replicated four times, were systematically distributed among 64 plots, each 0.5 × 0.5 m in size and 1.5 m apart. Four tubes, each 20 mm in diameter and 0.4 m long, were hammered into the ground at the plot corners to make it possible to relocate the plots using a four-legged frame.

elevation

The plots were established in two 5 × 35 m rectangular areas at different elevations (c. 15 m apart) orientated parallel to the lake edge. Each elevation area was divided into quarters. The remaining three factors (eight treatments) were systematically applied to each of the quarters, in both elevation areas. The lake water fluctuates seasonally and reaches its highest level in spring (March–April). The low-elevation plots were close to the spring high-water level, while the high-elevation plots were 0.5–0.6 m higher. To exclude larger animals from the experiment, both areas were enclosed within metal fences (mesh size: 20 × 20 mm), about 1.6 m high above ground and 0.4 m below ground.

vegetation removal

The vegetation in 16 plots at each elevation was removed completely. Care was taken to avoid disturbing the soil surface and the underground parts of the vegetation. To reduce edge effects, the plants in a 0.2-m belt outside the plots were cut and removed. During the growing seasons, we also regularly trimmed the surrounding plants rooted outside the plots to minimize shading.

seed addition

The seeds used in this experiment were all collected from within the Wicken Fen area in October 2000 and comprised 18 species (Table 1). These species were chosen because: (i) they occurred in the surrounding areas but were absent from or (in one case) rare in the study plots; (ii) they represented a range of life forms; and (iii) they were autumn seeding.

Table 1.  The species used for seed addition in the experiment and their emergence percentage recorded in the plots with bare ground and without litter
SpeciesLife form*Number of seeds sown per plotEmergence (%)
  • *

    Life form: w = woody; f = forb; g = graminoid.

  • Calculated as the means of the number of seedlings in the plots.

Alnus glutinosaw100 5.50
Angelica sylvestrisf10018.50
Betula pendulaw100 0.00
Calamagrostis epigejosg100 0.00
Cladium mariscusg100 0.00
Epilobium hirsutumf20056.75
Eupatorium cannabinumf10045.50
Filipendula ulmariaf10039.50
Iris pseudacorusf100 4.50
Juncus subnodulosusg200 2.83
Lysimachia vulgarisf10028.25
Lythrum salicariaf20047.75
Molinia caeruleag200 0.00
Pulicaria dysentericaf20035.50
Rhamnus catharticusw100 8.50
Scrophularia auriculataf20046.88
Sparganium erectumg100 0.00
Viburnum opulusw100 0.00

All seeds were brought back to the laboratory after collection, dried at room temperature and separated from inflorescences. Mixtures of 2400 seeds, 100 or 200 seeds per species, were put in nylon stockings and stratified in moist sand under 4 °C and darkness for more than 3 months. The seeds in each bag were evenly distributed by hand within a single ‘seeds added’ plot (n = 32) on 22 April 2001. To minimize seed predation by birds, the plots were covered with a bird-net during the first 2 months after sowing.

litter addition

Litter was collected from a nearby meadow site where the vegetation had been cut in the previous year and 2000 g m−2 of room-dried material, mostly composed of graminoid species, was added to each of 32 plots. Previous experiments had shown that this amount of litter influenced plant communities across Europe (Nilsson et al. 1999). It also exceeded the local above-ground biomass production, which was about 1200 g m−2. The litter was applied by hand over a 0.7 × 0.7 m2 square that included a buffer-strip of 0.2 m outside the plot, after gently removing previous litter. Litter, which was well shaken by hand to reduce its seed content, was applied after sowing the seeds, i.e. litter was put on top of the seeds wherever both litter and seeds were applied.

monitoring of groundwater levels

Soil water pressure or tension during the study period was measured by three pairs of tensiometers evenly spaced among the plots in each area, with one of each pair buried at 0.25-m depth and the other at 0.75-m depth. An additional tensiometer (acting as a pressure transducer) was located in the lake, c. 0.30 m below the 2001 spring water level, to measure the water stage.

biomass and soil analysis

To examine possible differences in production and soil nutrients between the two elevations, biomass was sampled at three evenly distributed places within each area in September 2001 and three topsoil (< 20 cm) samples were taken in April 2002. Biomass was sampled from an area 50 × 50 cm in size, and quantified as room-dry matter. The analysis of soil pH, inline image, inline image and K+ was carried out in the soil laboratory, Department of Geography, University of Cambridge. The inline image was extracted from the soil with saturated calcium sulphate solution, the phosphorus was extracted at about 20 °C with sodium bicarbonate solution at pH 8.5, and potassium was extracted with 1 m ammonium nitrate (Ministry of Agriculture, Fishery and Food 1986).

data collection and analysis

Seedling emergence and subsequent changes in seedling abundance were censused in each plot, using an aluminium frame 0.5 × 0.5 m in size. The four legs of the frame were inserted into the tubes at the corners of each plot. The presence of all vascular plants (including the seedlings) was surveyed between 5 and 14 September 2001 and between 29 July and 5 August 2002. The data were then analysed by the UNIANOVA method (GLM Univariate Analysis; SPSS Incorporated 2000), with species richness, total number of seedlings and number of individuals originating from sown seeds as dependent variables, and elevation, vegetation canopy, plant litter and seed availability as independent variables.

The frequency of the species originally present was expressed by the percentage of the 64 plots in which the species appeared (Table 2). To determine the relative contribution of individual factors and interactions to the variation in species richness and seedling emergence, we also used the ratio of sum of squares of examined factors (individual or interactive) in the UNIANOVA procedure to the total sum of squares (for all the factors, their interactions and errors).

Table 2.  The species originally present in the study plots
SpeciesLife form*Life spanRhizomatousStoloniferousFrequency (%)
  • *

    f = forb; g = graminoid.

  • a = annual; b = biennial; p = perennial.

  • Based on appearance in all 64 plots.

Cirsium arvensefpx 100.00
Elytrigia repensgp x100.00
Glechoma hederaceafp x 98.44
Agrostis stoloniferagpxx 87.50
Potentilla anserinafp x 82.81
Holcus lanatusgp   70.31
Arrhenatherum elatiusgp   62.50
Poa trivialisgp x 57.81
Potentilla erectafpx  53.13
Carex hirtagpx  51.56
Cerastium fontanumfp   43.75
Ranunculus repensfp x 37.50
Galium mollugofp x 32.81
Calystegia sepiumfpx  31.25
Sonchus arvensisfp   29.69
Cirsium vulgarefb   26.56
Juncus inflexusgpx  26.56
Festuca rubragpx  12.50
Geranium dissectumfa/b   12.50
Galium aparinefa   10.94
Persicaria maculosafa    7.81
Tussilago farfarafpx   6.25
Centaurea nigrafp    4.69
Plantago majorfp    4.69
Phragmites australisgpx   3.13
Prunella vulgarisfp x  3.13
Rumex spp.fp    3.13
Senecio jacobaeafb    3.13
Sinapis arvensisfa    3.13
Trifolium repensfp x  3.13
Urtica dioicafpx   3.13
Achillea millefoliumfpx   1.56
Aethusa cynapiumfa    1.56
Epilobium hirsutumfpx   1.56
Galium uliginosumfpxx  1.56
Lamium albumfp x  1.56
Reseda luteafp    1.56

Results

The tensiometer recordings showed that plot elevation was a good surrogate for moisture availability. In the low elevation plots, soil water potential ranged from −27.39 to −67.82 hPa at 25 cm and from −36.09 to −58.33 at 75 cm between June and September 2001. In the high elevation plots, however, soil water potential varied dramatically at 25 cm (−46.32 and −354.92 hPa), but remained quite stable (between −88.37 and −111.89 hPa) at 75 cm. During the experiment, no plots were ever flooded, but the low elevation plots were clearly less susceptible to drought stress than the high elevation plots, where the suctions required for extracting moisture from the soil were beyond field capacity (a suction of c. 100 hPa) for 57 of 85 days examined in 2001.

During the main growing season, from May to September 2001, monthly rainfall was generally below average, while monthly mean temperatures were higher than average for the last 30 years (recordings at Cambridge University Botanic Garden, 30 km from the study site).

The averages of above-ground biomass in September 2001 were 1180.7 ± 99.5 g m−2 at high elevation, and 1200.5 ± 73.2 g m−2 at low elevation. The soil chemistry analysis indicated that there were no significant differences in pH, inline image, inline image and K+ levels between the high and low elevation soils (anova, P > 0.05 for all, n = 6).

In total, 12 of the 18 sown species, most of them forbs, emerged in the field. Thirty-seven species were present in the community before sowing, of which only one, Epilobium hirsutum, was also sown (Table 2).

We concentrate here on the results from September 2001. By then, five sown species (Epilobium hirsutum, Eupatorium cannabinum, Lythrum salicaria, Pulicaria dysenterica and Scrophularia auriculata) had reached the stage of flowering and/or seeding. Individual plots contained 8–14 established species (i.e. those that were not sown in) and 0–11 seedling species (both from natural recruitment and from sowing), and a total of 5–263 seedlings had appeared, depending on the treatment (Table 3). Most established species were rhizomatous and/or stoloniferous and most seedlings originated from the sown seeds (Tables 2,3).

Table 3.  Species richness and seedling numbers (mean ± SE) for the 16 treatments in September 2001
TreatmentCode*Species richnessSeedling number
TotalEstablishedAll seedlingsSown speciesTotalSown
  • *

    E = elevation high (+) vs. low (–); V = vegetation intact (+) vs. removed (–); L = litter added (+) vs. not added (–); S = seeds added (+) vs. not added (–).

  • Originally present, not from sown seeds.

1E – V + L + S – 9.8 ± 0.5 9.8 ± 0.5 0 0  0  0
2E – V + L + S +11.0 ± 0.710.0 ± 0.7 1.3 ± 0.4 1.0 ± 0.0  9.5 ± 1.2  7.8 ± 1.1
3E – V + L – S –11.3 ± 0.511.3 ± 0.5 0 0  0  0
4E – V + L – S +13.3 ± 0.811.0 ± 0.7 2.5 ± 0.4 2.3 ± 0.5 12.8 ± 0.9 11.5 ± 1.1
5E – V – L + S – 9.5 ± 0.6 9.5 ± 0.6 0 0  0  0
6E – V – L + S +10.3 ± 0.4 8.8 ± 0.4 1.5 ± 0.5 1.5 ± 0.5 20.3 ± 1.2 20.3 ± 1.2
7E – V – L – S –11.0 ± 0.8 9.8 ± 0.9 1.3 ± 0.4 0  4.0 ± 0.8  0
8E – V – L – S +24.3 ± 0.913.8 ± 0.811.5 ± 0.710.5 ± 0.6264.5 ± 4.6262.3 ± 4.6
9E + V + L + S –13.0 ± 0.613.0 ± 0.6 0 0  0  0
10E + V + L + S +13.0 ± 0.511.5 ± 0.4 1.5 ± 0.4 1.5 ± 0.4 13.3 ± 1.5 13.3 ± 1.5
11E + V + L – S –12.3 ± 0.612.6 ± 0.6 0 0  0  0
12E + V + L – S +14.3 ± 0.612.8 ± 0.4 1.5 ± 0.5 1.5 ± 0.5  5.3 ± 0.8  5.3 ± 0.8
13E + V – L + S –11.3 ± 0.711.3 ± 0.7 0 0  0  0
14E + V – L + S +12.0 ± 0.610.0 ± 0.5 2.0 ± 0.5 2.0 ± 0.5 17.5 ± 1.5 17.5 ± 1.5
15E + V – L – S – 9.3 ± 0.6 8.5 ± 0.4 1.0 ± 0.5 0  1.8 ± 0.7  0
16E + V – L – S +14.8 ± 0.410.3 ± 0.5 4.5 ± 0.6 4.5 ± 0.6  9.8 ± 1.0  9.8 ± 1.0

relative importance of individual vs. interactive effects of factors

For all community parameters, a substantial proportion of the total variation could be attributed to interactive effects of the factors studied (Table 4). Individual factors were, however, more important for the seedling richness parameters, whereas two-way interactions dominated in the variables describing total species richness and seedling numbers.

Table 4.  Relative contribution (%)* of individual factors and their interactions to the total variation in community parameters. The four environmental factors are elevation (as a measure of soil moisture), vegetation canopy, plant litter, and seeds
SourceTotal species richnessRichness of all seedlingsRichness of seedlings from sown speciesTotal number of seedlingsNumber of seedlings of sown species
  • *

    Relative contribution (%) was calculated as the ratio of sum of squares of examined factors (individual or interactive) in the UNIANOVA procedure to the total sum of squares.

Individual factors30.5853.6352.3229.3228.80
Interactions between two factors39.2128.4629.4035.0135.17
Interactions between three factors11.9710.8812.0322.3222.68
Interactions between four factors 2.27 1.73 1.51 5.16 5.21
Errors15.98 5.30 4.75 8.19 8.14

effects of individual factors

Neither elevation nor vegetation canopy by itself affected total species richness (established + sown species) significantly, whereas litter (negative) and seed addition (positive) did (Table 5, Fig. 1). The richness of established species was significantly reduced by vegetation removal but not significantly affected by any other experimental factor (Fig. 1).

Table 5.  Results of the full factorial univariate analysis of variance on the five response variables. Total species richness = established species + seedling species from sown seeds
Source of variationd.f.Total species richnessRichness of all seedlingsRichness of seedlings from sown seedsTotal number of seedlingsNumber of seedlings of sown species
FPFPFPFPFP
E (elevation)1, 48 0.02    0.885 24.55< 0.001 18.24< 0.00136.96< 0.00135.55< 0.001
V (vegetation canopy)1, 48 1.71    0.198 98.18< 0.001 82.79< 0.00140.84< 0.00140.14< 0.001
L (litter addition)1, 4835.39< 0.001111.71< 0.001 89.69< 0.00130.02< 0.00128.70< 0.001
S (seed addition)1, 4854.76< 0.001251.35< 0.001337.97< 0.00164.09< 0.00165.51< 0.001
E × V1, 4818.95< 0.001 15.71< 0.001 15.21< 0.00134.88< 0.00135.14< 0.001
E × L1, 4827.28< 0.001 35.35< 0.001 33.14< 0.00137.52< 0.00137.10< 0.001
E × S1, 48 6.82    0.012 21.38< 0.001 18.24< 0.00135.70< 0.00135.55< 0.001
V × L1, 4812.13    0.001 79.53< 0.001 57.97< 0.00132.47< 0.00130.86< 0.001
V × S1, 4818.95< 0.001 48.11< 0.001 82.79< 0.00137.52< 0.00140.14< 0.001
L × S1, 4833.68< 0.001 57.71< 0.001 89.69< 0.00127.19< 0.00128.70< 0.001
E × V × L1, 4811.14    0.002 18.44< 0.001 15.21< 0.00131.43< 0.00130.73< 0.001
E × V × S1, 48 3.56    0.065 13.20    0.001 15.21< 0.00133.67< 0.00135.14< 0.001
E × L × S1, 48 3.56    0.065 31.53< 0.001 33.14< 0.00136.26< 0.00137.10< 0.001
V × L × S1, 4817.71< 0.001 35.35< 0.001 57.97< 0.00129.52< 0.00130.86< 0.001
E × V × L × S1, 48 6.82    0.012 15.71< 0.001 15.21< 0.00130.28< 0.00130.73< 0.001
Figure 1.

Effect of individual factors on species richness (left column) and number of seedlings (right column). Data stem from September 2001. Different letters (i.e. a, b) between two adjacent bars indicate a significant difference (anovaP < 0.05).

All four individual factors studied had significant effects on both the richness and number of seedlings irrespective of whether all species, or only sown species, were considered (Table 5). In general, seedling emergence and seedling species richness were enhanced by lower elevation, seed addition and vegetation removal, but suppressed by litter addition (Fig. 1).

effects of interactions between factors

All interactions had strong effects on all five response variables, except that total species richness was only marginally significantly affected by the interactions between elevation × vegetation × seed and elevation × litter × seed (Table 5). Below, we concentrate on the effects of two-factor interactions on the variables of primary interest, i.e. total species richness and seedling numbers from sown seeds.

Although vegetation removal individually had no effect, it increased total species richness by an average of 2.4 species per plot at the low elevation, but reduced it at the high elevation by about 1.3 species (Fig. 2a). The interaction between litter and elevation had a similar effect; litter addition resulted in a reduction on average of 4.8 species per plot at the low elevation, but at the high elevation produced almost the same species number as in plots without litter (Fig. 2b). The difference in species richness between litter-covered and litter-free plots was smaller under a vegetation canopy than in vegetation-free plots (1.1 vs. 4.1, Fig. 2c), although more species were always found in litter-free plots. The interactions between seed availability and the other three factors were quite predictable (Fig. 2d–f); sowing added more species at the low elevation (4.3 vs. 2.1), in plots where vegetation had been removed (5.1 vs. 1.3) and in the plots without litter (5.7 vs. 0.7). Interaction affected seedling numbers (Fig. 2g–l) and total species richness in similar ways.

Figure 2.

Interactive effects of two factors on species richness (left column) and number of seedlings (right column). Data (mean ± SE) stem from September 2001. Different letters (i.e. a, b) between two adjacent bars indicate a significant difference (anovaP < 0.05).

Data from August 2002 showed the same patterns for both species richness and seedling emergence, although the surviving individuals of sown species decreased considerably compared with 2001.

Discussion

importance of interactions

Although a complexity of interactions between different factors has been assumed in plant community theory (Grime 1979; Huston 1979; Tilman 1982; Keddy 1989), few experiments have combined more than two factors to test such effects. By using a factorial design, we have successfully shown that plant diversity and colonization in the studied community are considerably determined by interactive effects of ecological factors, as indicated by the fact that 41–63% of the total variation was explained by these interactions. Some individual factors may have a minor influence, for example the effects of elevation and vegetation canopy on total species richness, but their interactions with each other or with other factors, such as litter accumulation, may have a substantial influence on total species diversity and plant colonization. Moreover, the effects of interactions between factors may also be reversed in contrast to what would be expected from the effects of the individual factors (e.g. the interaction between litter and elevation). Consequently, knowledge about the role of individual factors only, and ignorance of their interactions, may lead to false predictions about community structure and function.

role of vegetation canopy vs. elevation

The fact that vegetation removal increased total species richness at the low elevation but reduced it at the high elevation suggests that species interactions (competition) were more intense at the low elevation than at the high elevation. During this study the soil water content at the high elevation was far below the field capacity for most of the critical period of seedling emergence and original species recruitment. Removal of vegetation canopy, i.e. a removal of the species interactions, in the high elevation plots apparently did not improve the microenvironment for species recruitment. In contrast, water supply was not limiting at the low elevation and competition from established vegetation probably prevented germination and colonization (cf. Grime 1979; Twolan-Strutt & Keddy 1996).

By means of biomass removal experiments, Brewer et al. (1997) and Lenssen et al. (1999) found that species richness along a flooding gradient was controlled by abiotic factors in frequently flooded zones, but by species interactions at higher elevations. In both studies, plants were exposed to flooding at low elevation, whereas our study involved no flooding at low elevation but dry soil conditions at high elevation. Consequently, competition may be weaker at both ends of a hydrological gradient in floodplain meadow communities.

role of litter accumulation vs. elevation

Previous studies have shown that accumulation of plant litter generally reduces species richness in most plant communities (Xiong & Nilsson 1999). Factors removing litter, e.g. flooding, may therefore be important in maintaining a species-rich flora in productive communities where litter usually accumulates. We therefore hypothesized that wet conditions in a riparian zone should alleviate effects of litter accumulation on the plant community.

The fact that litter addition reduced species richness at the low elevation and favoured the number of seedlings at the high elevation does not support the hypothesis above. Reduced species richness in the low elevation plots where litter was added suggests that regular flooding or other disturbances are needed to reduce negative impacts of litter in the long term. The positive effects of litter on seedling numbers at high elevation may have been due to the physical effect of preserving soil water (cf. Facelli & Pickett 1991; Xiong & Nilsson 1997).

Our results indicate that litter addition mainly influenced species colonization, consistent with earlier findings indicating that germination is more sensitive to litter accumulation than other stages of vegetation development (Xiong & Nilsson 1999).

effects of vegetation vs. litter addition

We found that the interaction between vegetation canopy and plant litter influenced plant colonization and species richness as in earlier studies (Foster & Gross 1997, 1998; Suding & Goldberg 1999). However, the relative importance of vegetation canopy and plant litter varied with the community parameter considered (species richness or number of seedlings), as well as with elevation and stage of vegetation development.

In general, the effect of plant litter by itself was more important than that of vegetation canopy in affecting total species richness and seedling species richness, but vegetation had a stronger individual effect than litter on seedling numbers (Table 5). This perhaps reflects a major difference in that the added litter might act as a filter by allowing only certain species to germinate and emerge through the litter layer (cf. Facelli & Pickett 1991; Xiong et al. 2001), whereas canopy shade, which also affects both germination and seedling survival, may discriminate less between species (Tilman 1993; Suding & Goldberg 1999; Jutila & Grace 2002). The weaker effect of litter on richness and seedlings of the sown species in the vegetation-intact than in the vegetation-removed plots (Fig. 2c,i) indicates that the stage of development may also be important. A dense canopy excludes most potential colonisers by effectively blocking light from reaching the ground and the residual effect of litter is negligible. On bare ground, however, where almost no species interactions occur (Huston 1979; Pickett & White 1985), litter may play a critical role.

The relative importance of vegetation canopy and plant litter also changed with elevation. At low elevation, the effects of litter and vegetation on both species richness and seedling numbers were negative. As indicated by a separate analysis, litter had a stronger effect than the vegetation canopy on total species richness (anova, F = 44.1 vs. 11.3, P = < 0.001 vs. 0.003), whereas their effects on numbers of seedlings from the sown seeds were similar (P < 0.001 for both). It is likely that the less stressful microenvironment here resulted in more intense species competition that suppressed seedlings (Grime 1979; Twolan-Strutt & Keddy 1996). At the high elevation, on the other hand, vegetation canopy and litter both appeared to have more or less positive effects (Fig. 2a–b,g–h) and the vegetation canopy was more important than the plant litter in determining total species richness (P = 0.010 vs. 0.509), although its effect was unimportant in affecting numbers of seedlings from sown seeds, while litter addition increased the seedling numbers significantly.

importance of seed addition

Seed addition was the most important factor behind variation in species richness between the treatments in this experiment, which suggests that the community examined is seed-limited. The fact that species richness increased in almost all plots where seeds were introduced, even where litter and vegetation were present, shows that suitable regeneration niches (Grubb 1977; Turnbull et al. 2000) were available. Thus, the community was not species-saturated and is likely to be enriched by introduction or immigration of species (Cornell & Lawton 1992; Zobel et al. 2000; Levine 2001).

After removal of both vegetation and litter, the total numbers of seedlings were much lower in plots without added seeds than in those where seeds had been added; thus creating bare ground or gaps by vegetation and litter removal did not result in as rich a species community as did seed addition. This agrees with findings that local seed banks have a limited importance for species diversity, for example, in woodlands (Eriksson & Ehrlén 1992), calcareous grassland (Zobel et al. 2000), and wetland communities (Brown 1998; Levine 2001; Smith et al. 2002). Although water conditions, species interactions and litter accumulation are still limiting factors, dispersal limitation and perhaps poor seed viability may well be reasons for the previous absence of species that were established from seeds sown in this experiment (cf. Cornell & Lawton 1992; Levine 2001).

Acknowledgements

We thank Dr Owen Mountford for advice in selecting and collecting the seed species, Alex Goodall at the Botanic Garden of The University of Cambridge for helpful advice and daily care of the germination trials, Magnus Svedmark and Elisabet Carlborg for preparing seeds, Melanie Edmunds and Helen Peeks for assistance in the field, and Chris Rolfe for soil analysis. We gratefully acknowledge the comments of the editors, Chris Kellogg and an anonymous reviewer. Thanks also go to the staff and the Management Committee of Wicken Fen National Nature Reserve for providing the study site and other support. This work was financed by the European Commission Environment Programme under Research Grant CT-1999–00031 FLOBAR2. We are grateful to Dr H. Barth for his interest in the project.

Ancillary