Does pattern of soil resource heterogeneity determine plant community structure? An experimental investigation

Authors

  • DUSHYANTHA K. WIJESINGHE,

    1. Department of Biology and Environmental Science, John Maynard Smith Building, School of Life Sciences, University of Sussex, Falmer, Brighton, East Sussex BN1 9QG, UK
    Search for more papers by this author
  • ELIZABETH A. JOHN,

    Corresponding author
    1. Department of Biology and Environmental Science, John Maynard Smith Building, School of Life Sciences, University of Sussex, Falmer, Brighton, East Sussex BN1 9QG, UK
    Search for more papers by this author
  • MICHAEL J. HUTCHINGS

    1. Department of Biology and Environmental Science, John Maynard Smith Building, School of Life Sciences, University of Sussex, Falmer, Brighton, East Sussex BN1 9QG, UK
    Search for more papers by this author

Elizabeth A. John (tel. +1273 877256; fax +1273 678433; e-mail e.a.john@sussex.ac.uk).

Summary

  • 1Spatial and temporal pattern of nutrient delivery to individual plants and plant populations can affect growth and allocation of biomass to roots and shoots. We predicted that it would also affect attributes of plant community structure, including species composition, diversity, and partitioning of biomass between above- and below-ground parts. We tested these predictions experimentally by providing the same quantity of nutrients in five different patterns to sown plant communities grown under field conditions for 2 years. We used orthogonal contrasts to compare the effects on community structure of homogeneous vs. heterogeneous nutrient supply, supply of nutrients in patches at two different densities, and provision of nutrients in spatially predictable vs. unpredictable patches. We harvested above- and below-ground plant biomass, and measured species richness, diversity, and species population sizes.
  • 2Pattern of nutrient delivery significantly affected community biomass and below-ground : above-ground biomass partitioning. Treatments with the lowest density of nutrient-rich patches supported 44% more total biomass than homogeneous treatments, largely as a result of greater (71% more) below-ground biomass, which resulted in a 39% increase in community below-ground : above-ground biomass ratio.
  • 3Above-ground community composition was also affected by treatment. There were several instances in which biomass, percentage cover or population sizes of different species, were affected by treatment. There were few significant differences between communities in treatments with predictable and unpredictable nutrient patches. Treatment did not affect species richness or diversity, despite high power to detect small differences.
  • 4We conclude that the spatial and temporal pattern of nutrient supply strongly affects some important facets of plant community structure, but has less influence on others. Further understanding of community responses to the pattern of nutrient supply will require experiments testing the responses of individual species to heterogeneity with and without competitors present.

Introduction

Soil-based resources required for plant growth are heterogeneously distributed at a variety of scales in time and space (e.g. Bell & Lechowicz 1991; Jackson & Caldwell 1993; Kleb & Wilson 1997). It is now recognized that this is a ubiquitous feature of natural ecosystems with significant ecological consequences (Casper et al. 2000; Fitter et al. 2000; Wilson 2000). For example, recent studies have shown that the pattern of provision of soil-based resources can significantly affect the performance of individual plants and plant populations. Growth of individual plants can be increased by up to fourfold by different patterns of nutrient supply (Birch & Hutchings 1994; Wijesinghe & Hutchings 1997, 1999; Einsmann et al. 1999), and the yield of monocultures by as much as 40%, depending on the spatial pattern in which a fixed quantity of nutrients is supplied (Day et al. 2003a). Responses to the spatial pattern of nutrient delivery are both species and pattern specific (Einsmann et al. 1999; Farley & Fitter 1999; Wijesinghe et al. 2001).

Many plant species proliferate roots in nutrient-rich patches (Robinson 1994 and references therein), but species differ both in their ability to select such patches and in the speed with which they can take up the available nutrients (Campbell et al. 1991; Robinson 1994; Einsmann et al. 1999; Farley & Fitter 1999; Fransen et al. 1999; Robinson et al. 1999; Hutchings et al. 2000; Wijesinghe et al. 2001). As rapid occupation of a nutrient-rich patch may allow a plant to pre-empt nutrients from a competitor (Robinson et al. 1999), it is not surprising that heterogeneity in resource supply can alter competitive hierarchies among plant species (Fitter 1982; Fransen et al. 1999; Novoplansky & Goldberg 2001), and increase the intensity of competition experienced by plants (Day et al. 2003c), even though heterogeneity does not always affect the outcome of competition (Cahill & Casper 1999).

A recent study by Gersani et al. (2001) has important implications for our understanding of plant responses to heterogeneity. Soybean (Glycine max) plants sharing a resource supply with one other plant produced more root biomass and less seed biomass than plants grown alone with access to half as much resource. This result satisfies the prediction of the Tragedy of the Commons model (Hardin 1968) that individuals competing for a resource to which both have free access will invest more into acquiring that resource than they would if it were not accessible to both. Thus, plants competing for soil-based resources should devote a higher proportion of their biomass to roots than non-competing individuals provided with the same per capita resource supply. We predict that this effect will be exacerbated when nutrients are heterogeneously distributed, because the nutrients (and therefore, in many cases, plant roots) occupy a smaller volume of substrate. If this is true, communities of plants competing for patchily distributed soil resources will devote a greater proportion of their biomass to roots than communities of plants competing in homogeneous conditions.

Plant species differ in their ability to acquire patchily distributed nutrients (Campbell et al. 1991; Einsmann et al. 1999; Wijesinghe et al. 2001). For instance, when growing alone, plants with smaller root systems are more disadvantaged by starting life a given distance from a nutrient source (Wijesinghe et al. 2001; Hutchings et al. 2003). This effect is likely to be exacerbated in a community context where many plants are competing for resources. Furthermore, the most positive responses to spatial heterogeneity occur when there is a match between the scale of a plant's root system and the scale of patches. Heterogeneity at other scales may be less beneficial or even detrimental to growth, as plants struggle to achieve the optimal distribution of resource acquiring structures for the scale of the patches (Wijesinghe & Hutchings 1997; Wijesinghe et al. 2001). Different species might thus be expected to respond idiosyncratically to spatial heterogeneity (and probably also to temporal heterogeneity; see, for example, Novoplansky & Goldberg 2001). Thus, not only the presence of heterogeneity, but also its precise configuration could be expected to have implications for the relative success of individual species and for community structure.

If heterogeneity in resource supply has an impact on community structure, it will also affect the ability of newly arriving species to colonize the community, because of different availabilities of suitable colonization patches in communities with different patterns of resource supply. This could arise because of differences in overall biomass and overall intensity of competition within the community, or because some parts of the habitat support less biomass than others. Alternatively, if the position and time of origin of new resource-rich patches is unpredictable, the existing community may be unable to occupy new patches before colonizing species establish (Davis et al. 2000; Davis & Pelsor 2001).

Crick & Grime (1987) proposed that species that forage for resources primarily by morphological means will be more successful in habitats where resource availability is high and relatively predictable, whereas species that forage primarily by physiological means will be more successful in habitats where resource availability is low and patches of resource abundance are unpredictable in space and time. The fact that species differ in the extent to which their roots explore the substrate rather than remaining sedentary could be reflected in differences in community structure between habitats where the same locations are repeatedly enriched with resources and habitats where enrichment is spatially unpredictable through time. Species with extensive and rapid root growth should be at an advantage in situations where the supply of resources is spatially unpredictable.

Few experimental studies have examined the effects of resource heterogeneity on plant community structure. Such studies often confound resource heterogeneity and total resource availability (Collins & Wein 1998a,b), although a recent exception is the work by Stevens & Carson (2002) in which light heterogeneity was not found to control diversity. Descriptive (non-experimental) studies commonly report a positive correlation between the degree of environmental heterogeneity and species richness (e.g. Reynolds et al. 1997; Sulkava & Huhta 1998; Huston 1999; Therriault & Kolasa 2000; for an experimental example see Vivian-Smith 1997). This pattern can be explained by models in which niche differentiation allows coexistence in heterogeneous habitats (Fitter 1982; Tilman 1988; Pacala & Tilman 1994; Huston 1999), because different species dominate patches of different quality. In most such models, single plants occupy single patches. However, when the scale of heterogeneity is smaller than that of individual organisms, single plants of species that are successful in heterogeneous environments may be able to dominate both rich and poor patches, excluding less successful species from poor-quality patches that could otherwise act as refuges from competition (Day et al. 2003b). This could result in reduced species richness under heterogeneous conditions, and might help to explain some published results. For instance, Kleb & Wilson (1997) recorded a negative correlation between soil resource heterogeneity and plant species richness in natural vegetation, and Collins & Wein (1998a) found no effect of scale of heterogeneity on species richness. In order to understand the effect of heterogeneity on community structure and species richness, experimental treatments are needed that only differ in the pattern of resource supply to the community.

We report such an experiment here. Artificial communities were established from a standard seed inoculum containing species with a wide range of ecological strategies, although all were typical of roadside verges in the UK. Nutrients were supplied either homogeneously or in one of four heterogeneous patterns. The four heterogeneous treatments differed in the density of nutrient-rich patches and in whether the same or different patches of substrate were enriched at consecutive feeding dates. Treatments with lower densities of nutrient-rich patches were enriched more frequently, so that the same total nutrient supply was delivered in all treatments, including the homogeneous treatment. Planned orthogonal contrasts were used to determine the effects on community structure and species richness of homogeneity vs. heterogeneity, the effects of density of nutrient patches, and the effects of consistency or randomness in patches of nutrient enrichment. We examined the effects of pattern of nutrient supply on community biomass, below- and above-ground biomass partitioning, and on species composition (species richness, diversity and relative abundance of all species within the community).

We made the following predictions:

  • 1The contrast between homogeneous and heterogeneous treatments would reveal strong effects on community structure. (i) Communities growing in heterogeneous environments would have greater overall biomass than those in the homogeneous treatment. This would be caused primarily by an increase in root biomass, resulting in higher community below-ground : above-ground biomass ratios in the heterogeneous treatments. (ii) Due to more intense below-ground competition, communities in the heterogeneous treatments would have lower diversity and species richness than the community in the homogeneous treatment. (iii) Changes in the competitive hierarchy between the homogeneous and heterogeneous treatments would lead to differences in relative abundance of species and in overall community composition. (iv) Because of their predicted greater total biomass, the communities in heterogeneous treatments would also be less susceptible than communities in homogeneous environments to invasion by species not included in the original inoculum.
  • 2The contrast between the heterogeneous treatments with different densities of nutrient patches would reveal differences in community structure. This would be caused by differences in the ability of species to acquire nutrients in environments differing in the spacing and temporal frequency of nutrient supply. This contrast should primarily reveal differences in community composition, because the performance of different species will be altered to different extents by changes in the mean distance between patches of nutrient enrichment.
  • 3The contrast between treatments in which the same locations were repeatedly enriched with nutrients, and those in which patches of enrichment were selected at random at each feeding occasion, would produce differences in community structure.

This contrast should primarily reflect differences in the ability of species to acquire resources in spatially unpredictable environments.

Materials and methods

The mixture sown consisted of 20 herbaceous annual and perennial species with a range of life-history strategies (Table 1). Each experimental replicate was sown with an inoculum containing approximately 125 seeds of each of the 20 species (2500 seeds in total) mixed with 300 mL of potting sand for even sowing. Seeds were sown in situ onto the surface of the substrate in 120 experimental boxes placed in the University of Sussex field trials plot at Falmer on 20–21 April 1998. The substrate was kept moist to encourage germination and establishment. The communities were allowed to develop over two full growing seasons. Non-sown species that established in the boxes during this time were not weeded.

Table 1.  Species used in the seed inoculum, their plant families and whether they are annual (A) or perennial (P). Nomenclature follows Stace (1991). All species are common constituents of British road verge communities. Mean rank abundance (across 24 replicates) at the final harvest is shown for each species in each treatment (see Methods for details of treatments),with rare (r) and absent (a) species identified. Colonist (i.e. non-sown) species present at the final harvest were Achillea millefolium, Cerastium fontanum, C. holosteoides, Convolvulus arvensis, Epilobium hirsutum, Geranium molle, Geranium pyrenaicum, Holcus lanatus, Juncus spp., Lolium perenne, Prunella vulgaris, Rubus fruticosus agg., Rumex acetosella, Taraxacum spp., Trifolium repens, unidentified moss, unidentified seedlings
SpeciesFamilyAnnual/ perennialHOMHET50PHET50UHET25PHET25U
Agrostis stoloniferaPoaceaeP3.73.84.13.93.6
Anagallis arvensisPrimulaceaeAaaaaa
Arrhenatherum elatiusPoaceaeP2.52.32.52.52.3
Campanula rotundifoliaCampanulaceaePrrrrr
Capsella bursa-pastorisBrassicaceaeAaaaaa
Centaurea nigraAsteraceaePaaaaa
Dactylis glomerataPoaceaeP2.11.72.31.71.9
Daucus carotaApiaceaeP6.77.47.77.28.0
Festuca ovinaPoaceaeP5.55.55.45.05.4
Galium verumRubiaceaeP9.19.09.28.28.5
Lapsana communisAsteraceaeAaaaaa
Leontodon hispidusAsteraceaePrrrrr
Medicago lupulinaFabaceaeP6.26.75.46.36.0
Papaver rhoeasPapaveraceaePaaaaa
Plantago lanceolataPlantaginaceaeP4.44.03.83.63.2
Poa annuaPoaceaeA/Praaaa
Ranunculus acrisRanunculaceaePrrrrr
Rumex crispusPolygonaceaeP8.38.28.38.58.5
Sonchus oleraceusAsteraceaeAaaaaa
Torilis japonicaApiaceaeAaaaaa
Rare species  9.69.79.39.98.9
Colonist species  8.18.28.19.79.9

The boxes were 50 × 50 × 10 cm in size. They had wooden bases with drainage holes. The boxes were sunk into holes of the correct size dug into the field site, and filled with substrate until the surface was flush with the surrounding soil. The substrate in each box consisted of 20 L of a 3 : 1 mixture of sterilized sand and John Innes potting compost.

The surface of each box was divided into 64 6.25 × 6.25 cm patches for nutrient addition and recording purposes. There were no partitions between these patches, enabling roots to grow between them. During each of two growing seasons, each box received 64 Phostrogen (Solaris, Monsanto, Buckinghamshire, UK) plant growth tablets, containing NPK in an 8 : 11 : 23 ratio, with a 30-day release time. The spatial and temporal patterns in which the tablets were supplied constituted the five treatments, and all treatments received the same quantity of nutrients within each growing season. In the homogeneous treatment (HOM), every patch received one tablet at the beginning of each growing season. The four heterogeneous treatments were provided with nutrient tablets at one of two densities. In the HET50 treatments half the patches received a Phostrogen tablet at each of two dates during each growing season. The 32 enriched patches were selected randomly for each replicate at the first date. At the second date, the predictable treatments (HET50P) were fed again in the same locations, whereas 32 of the 64 patches were again selected at random in the unpredictable treatment (HET50U). In the HET25 treatments one-quarter of the patches received tablets at each of four dates throughout each growing season. The feeding protocol was similar to that for the HET50 treatments, but the 16 patches used for the first enrichment were selected randomly for each replicate and subsequent enrichment either occurred in the same 16 patches on each occasion (HET25P), or in different randomly selected patches on each occasion (HET25U). Randomization of the patches to be fed with nutrients was carried out independently for every replicate in the four heterogeneous treatments.

The first nutrient addition for all treatments was between 22 and 24 April 1998. Subsequent dates of enrichment were 1 July 1998 for the HET50 treatments and 27 May, 1 July and 3 August 1998 for the HET25 treatments. In 1999 the first enrichment was between 16 and 18 March, with subsequent enrichment on 1 June for the HET50 treatments and on 21–28 April, 1 June and 2 July for the HET25 treatments.

Treatments were distributed in the field in a randomized block arrangement, with four blocks, and six replicates per treatment per block (i.e. 24 replicates per treatment in total).

sampling and harvesting regime

The communities were surveyed non-destructively twice in 1998 (first survey 29 June to 8 July, second survey 30 August to 8 September), and once in 1999 (1–11 June). For each survey, a pin was placed at the centre of each of the 64 patches in each box, and all touches of the pin by any species were recorded. A destructive harvest of the whole experiment, carried out block by block, was conducted during August 1999. Above-ground plant parts were cut at ground level, sorted to species, bagged, oven dried and weighed. The number of individual plants (genets) of each species in each replicate was recorded. Below-ground parts of plants were carefully washed free of substrate, bagged, oven dried and weighed. Separation of roots of different species was not possible.

analysis

Data were log-transformed where necessary to meet the assumptions of anova. As this was a community experiment and the performance of different species could not be considered to be independent of each other, manovas were used initially to look for overall treatment effects. When appropriate these were followed by univariate anovas (with four treatment degrees of freedom and three block degrees of freedom), which were followed by planned orthogonal contrasts (Sokal & Rohlf 1981). These were used to compare the effects of heterogeneity against homogeneity (HOM vs. HET50P, HET50U, HET25P and HET25U), the effects of the density of nutrient-rich patches in the heterogeneous treatments (HET25P and HET25U vs. HET50P and HET50U), and the effects of predictability against unpredictability in positions of nutrient supply (HET50P vs. HET50U and HET25P vs. HET25U). As this was a mixed-model design, treatment effects were tested over the block–treatment interaction. Discriminant function analysis (based on the manovas) allowed us to investigate the distinctiveness of the communities produced by the treatments.

For analysis, species were considered individually (for dominant species), or in one of two groups referred to as rare species and colonists. Dominant species were defined as those that individually contributed at least 10% to the community biomass at the final harvest. Rare species were defined as species included in the inoculum that were found in less than half of the replicates of each treatment at final harvest. Colonists were those species that were not in the inoculum but which had established in the boxes during the experiment.

The following variables were analysed: percentage cover of each of the main species and species groups, species richness (number of species), Shannon index of diversity (Magurran 1988), calculated from the number of individual plants, evenness, above-ground biomass of each species or group of species (as defined above), relative abundance (proportion of total above-ground biomass) of each species or group of species, community below-ground : above-ground biomass ratio, and population size (number of individual plants) of each species.

Results

community biomass

There was a highly significant effect of treatment on the total biomass of the community at final harvest (Table 2, Fig. 1), with 44% more biomass in the most productive treatment (HET25U) than in the least productive (HOM). The contrasts between homogeneous and heterogeneous treatments and between 25% and 50% enrichment treatments were significant, but there was no significant effect on community biomass as a result of enrichment being predictable or unpredictable in space. The increase in total biomass from the least to the most productive treatment reflected increases in both below- and above-ground biomass (Fig. 1), but the percentage increase in below-ground biomass (71%) was greater and of high statistical significance whereas the 20% increase in above-ground biomass was not significant. As a result of the differential responses of below- and above-ground biomass to pattern of nutrient delivery, there were large differences in community below-ground : above-ground biomass ratios between treatments, with significant effects in the homogeneous vs. heterogeneous contrasts. The below-ground : above-ground biomass ratio for the HET25U treatment, in which the ratio was highest, was 39% greater than in the HOM treatment, in which the ratio was lowest (Fig. 1). Despite the relatively shallow pots used in the experiment, there was no obvious evidence of root binding in the treatments by the final harvest.

Table 2.  Effects of treatment on community biomass. Results of mixed-model anovas of log-transformed data. All main effects are shown, together with those planned orthogonal contrasts that produced a significant result, for total biomass, root and shoot biomass and community below-ground : above-ground biomass ratio. Prior to conducting univariate anovas, manova was performed on the shoot and root data (treatment effect: Wilk's Lambda = 0.696, F = 5.507, d.f. = 8, 22, P = 0.001). Treatment codes: HOM = all patches fertilized at same time; HET = only some patches fertilized; HET50 and HET25 = 50% or 25% of patches, respectively, fertilized at each feeding
VariableSourced.f.Mean squareF-ratioP
Total biomassTreatment  40.416 9.282< 0.001
HOM vs. HET  11.15719.95< 0.002
HET50 vs. HET25  10.437 9.747< 0.05
Block  31.26728.243< 0.001
Treatment × Block 120.058 1.334NS
Error1000.045  
Above-ground biomassTreatment  40.129 2.362NS
Block  30.136 3.011< 0.05
Treatment × Block 120.055 1.222NS
Error1000.045  
Below-ground biomassTreatment  40.805 7.435< 0.01
HOM vs. HET  12.27921.102< 0.001
HET50 vs. HET25  10.892  8.259< 0.02
Block  33.54551.190< 0.001
Treatment × Block 120.108 1.565NS
Error1000.069  
Below-ground : above-ground biomass ratioTreatment  40.327 2.920NS
HOM vs. HET  10.798 7.125< 0.02
Block  32.65243.475< 0.001
Treatment × Block 120.112 1.836NS
Error1000.061  
Figure 1.

Mean (± SE) total community biomass, mean above-ground and below-ground community biomass and mean community below-ground :above-ground biomass ratio at final harvest. See Methods for description of treatments, and results of planned contrasts (Table 2) for significant differences between means and groups of means. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

species richness and diversity

Although there were some strong effects of treatment on individual species (see below), there were no effects on species richness in either year of the study, or on diversity or evenness at final harvest (Fig. 2a,b). Species richness declined equally in all treatments between the first and second years of the experiment.

Figure 2.

(a) Mean (± SE) species richness in the five treatments at the three survey dates and at final harvest. (b) Mean (± SE) Shannon indices of diversity and evenness in the five treatments based on numbers of plants at the final harvest. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

biomass of individual species and groups of species at the final harvest

manova performed using the biomass of the nine most abundant individual species, and summed biomass for the combined species groups of rare species and colonist species, demonstrated a significant overall treatment effect on above-ground biomass of individual species (treatment effect: Wilk's Lambda = 0.495, F = 1.799, d.f. = 44, 392, P = 0.002), with the discriminant function analysis revealing significant treatment-specific differences in the overall composition of the harvested communities (Fig. 3). For all treatments (in particular the HOM and HET25U treatments), discriminant function analysis placed more replicates in their own treatment group than in any other group (Table 3). However, in all treatments, between 10 and 16 of the 24 replicates were assigned to one of the other four groups. Thus, although treatment significantly affected community composition, some of the communities produced were not easily distinguished. There was, nevertheless, a clear directional shift in community composition from the HOM to the HET50 and HET25 treatments, with little overlap of groups between these treatments.

Figure 3.

Positions of plant communities harvested from individual replicates in August 1999, on axes defined by the first and second discriminant functions in a discriminant function analysis. The distribution of replicates between discriminant function groups (group numbers 1–5) is shown in Table 3. The canonical variables for each of these axes are as follows: Z1 = 0.055 Agrostis+ 0.0334 Arrhenatherum+ 0.067 Dactylis + 0.007 Daucus + 0.432 Festuca + 0.588 Galium– 0.006 Medicago + 0.339 Plantago – 0.469 Rumex – 0.109 rare species – 0.213 colonists. Z2 = 0.296 Agrostis + 0.014 Arrhenatherum+ 0.160 Dactylis + 0.209 Daucus + 0.528 Festuca – 0.244 Galium – 0.148 Medicago – 0.196 Plantago – 0.348 Rumex – 0.033 rare species + 0.802 colonists. The symbols show the treatment regimes applied to the individual communities; ◆ = HOM (all patches fertilized at the beginning of each growing season); □ = HET50P (the same 50% of patches fertilized on two occasions in each growing season); ▴ = HET50U (different 50% of patches fertilized on each of two occasions in each growing season); inline image = HET25P (the same 25% of patches fertilized on each of four occasions in each growing season); inline image = HET25U (different 25% of patches fertilized on each of four occasions in each growing season).

Table 3.  Distribution of treatment replicates between discriminant function groupings based on biomass of individual species at final harvest. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding
Predicted groupTreatment group
HOMHET50PHET50UHET25PHET25U
11446 1 0
2 484 4 0
3 569 3 7
4 12310 4
5 042 613

There were significant effects of treatment and at least one of the contrasts on the above-ground biomass of just two individual species, Dactylis glomerata and Galium verum, and strong but non-significant treatment effects for two other sown species (Plantago lanceolata and Rumex crispus) and for colonist species (Table 4, Fig. 4). Dactylisglomerata, G. verum and P. lanceolata produced more biomass in the HET treatments than in the HOM treatment, while the converse was true for R. crispus. Galiumverum produced significantly more above-ground biomass in the HET25 treatments than in the HET50 treatments. None of the individual species or species groups was significantly affected by the predictability of patches of enrichment. However, colonist species showed a strong response both to density of nutrient-rich patches, and to predictability at the HET50 scale, with over four times as much above-ground biomass being produced in HET50P as in HET50U (Fig. 4).

Table 4.  Effects of treatment on above-ground biomass of individual species at the final harvest. Results of anovas of log-transformed data, including planned orthogonal contrasts (where significant), are shown for species for which the univariate anova showed a significant treatment effect. Prior to conducting univariate anovas, manova was performed using the biomass of the nine most abundant individual species (referred to as dominant species, see text), and summed biomass for the combined species groups of rare species and colonist species (treatment effect: Wilk's Lambda = 0.495, F = 1.799, d.f. = 44, 392, P = 0.002). Treatment codes: HOM = all patches fertilized at the same time; HET = only some patches fertilized; HET50 and HET25 = 50% or 25% of patches, respectively, fertilized at each feeding
SpeciesSourced.f.Mean squareF-ratioP
Dactylis glomerataTreatment  42596.470 2.739< 0.02
HOM vs. HET  19114.38915.070< 0.005
Block  3 747.001 0.755NS
Block × Treatment 12 604.818 0.612NS
Error100 988.978  
Galium verumTreatment  4   0.627 4.471< 0.02
HET50 vs. HET25  1   1.63711.693< 0.02
Block  3   0.220 1.821NS
Block × Treatment 12   0.140 1.160NS
Error100   
Figure 4.

Mean (± SE) biomass at final harvest in each treatment for species and groups of species for which there was a significant or notable overall treatment effect (Table 4). (a) Dactylis glomerata and Plantago lanceolata; (b) Galium verum and Rumex crispus; (c) colonist species. Results of planned contrasts are shown in Table 4 for significant differences between means and groups of means. See Methods for details of treatments. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

relative abundance

Relative abundance (based on above-ground biomass) only showed significant treatment effects for Galium verum, being higher in heterogeneous treatments than in the homogeneous treatment, and nearly four times greater in HET25 treatments than in HET50 treatments (Table 5, Fig. 5). Although not significant, due to the overriding effect of the strong block–treatment interaction term, Rumex crispus and colonist species showed much higher relative abundance in homogeneous than heterogeneous treatments (Fig. 5).

Table 5.  Effects of treatments on relative abundance of Galium, the only species for which the univariate anova showed a significant treatment effect, based on above-ground biomass at the final harvest. Results of anovas of angular-transformed data, including significant planned orthogonal contrasts, are shown. Prior to conducting univariate anovas, manova was performed using the angular-transformed relative abundances of the nine most abundant individual species (referred to as dominant species, see text), and the summed relative abundances of the rare and colonist species groups (treatment effect: Wilk's Lambda = 0.697, F = 0.816, d.f. = 44, 9, P = 0.697). Treatment codes: HOM = all patches fertilized at the same time; HET = only some patches fertilized; HET50 and HET25 = 50% or 25% of patches, respectively, fertilized at each feeding
SpeciesSourced.f.Mean squareF-ratioP
Galium verumTreatment  40.002 5.031< 0.02
HOM vs. HET  10.00410.00< 0.02
HET50 vs. HET25  10.00512.5< 0.01
Block  30.001 2.523NS
Treatment × Block 120.0004 1.220NS
Error1000.000  
Figure 5.

Mean relative abundance (± SE) at final harvest, of species for which there was a significant or notable overall treatment effect (Table 5). (a) Colonist species, (b) Galium verum, (c) Rumex crispus. Results of planned contrasts are shown in Table 5 for significant differences between means and groups of means. See Methods for details of treatments. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

population sizes

The manova on mean population size (number of individual plants of each species or group of species) was significant (Table 6), with the discriminant function analysis also revealing overall community effects (data not shown). Specific contrasts had significant effects on the population sizes of Festuca ovina, Galium verum and Plantago lanceolata. For these three species, populations were significantly larger in the heterogeneous treatments than in the homogeneous treatment (Fig. 6). For G. verum, there were also significantly more plants in HET25 treatments than in HET50 treatments. This contrast was also striking (although non-significant) for the colonist species, but they exhibited the opposite response to that of G. verum. There was no significant effect of predictability of patches of enrichment on number of plants for any species or group of species.

Table 6.  Analysis of population size data (numbers of plants per species) at final harvest. Results of anovas, including planned orthogonal contrasts, are shown for species for which the univariate anova showed a significant treatment effect. Prior to conducting univariate anovas, manova was performed (treatment effect: manova Wilk's lambda = 0.425, F = 2.233, d.f. = 44, 9, P = 0.047). Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding
SpeciesSourced.f.Mean squareF-ratioP
Dactylis glomerataTreatment  4 28.729 3.690  < 0.05
HET50 vs. HET25  1 61.760 7.933  < 0.05
Block  3 58.764 3.095NS
Treatment × Block 12  7.785 0.410NS
Error100 18.985  
Festuca ovinaTreatment  4171.554 9.591  < 0.001
HOM vs. HET  1537.63330.055  < 0.001
HET25P vs. HET25U  1117.187 6.551  < 0.05
Block  3136.811 2.662NS
Treatment × Block 12 17.888 0.348NS
Error100 51.400  
Galium verumTreatment  4 33.283 8.273<< 0.001
HOM vs. HET  1 69.008 9.526  < 0.02
HET50 vs. HET25  1 63.375 8.749  < 0.05
Block  3 23.022 5.722  < 0.001
Treatment × Block 12  7.244 1.801NS
Error100  4.023  
Figure 6.

Mean number of individuals per plot (± SE) at the final harvest for all species (a) and for individual species (b) for which there was a significant or notable overall treatment effect. Results of planned contrasts are shown in Table 6 for significant differences between means and groups of means. See Methods for details of treatments. Treatment codes: HOM = all patches fertilized at the same time; HET50P and HET50U = 50% of patches fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

percentage cover

Six species and the group of rare species showed significant effects of treatment on percentage cover in at least one of the 1998 surveys (Table 7, Fig. 7). At the first survey there was a tendency for the highest percentage cover for each species to be recorded from the HET25 treatments, with the HET50 treatments showing the lowest percentage cover (Fig. 7). Exceptions to this were Plantago lanceolata and Sonchus oleraceus, which both had their highest percentage cover in the homogeneous treatment. By the second survey in 1998, the lowest percentage cover values tended to be recorded in the HOM treatment, and the highest values in the HET25 treatments (Fig. 7), except for colonist species, which showed higher mean percentage cover values in the HET50 treatments. There was no overall effect of treatment on the early 1999 cover data, and only weak effects on individual species (Table 7, Fig. 7).

Table 7.  Analysis of percentage cover data from early and late 1998 (there were no significant effects in 1999). The contrasts between the predictable and unpredictable treatments were not significant for any species or species group. Probabilities shown under overall treatment effect are those associated with F-tests with d.f. = 3, 12. Probabilities shown for individual species in the contrast columns are those associated with F-tests with d.f. = 1, 12. Results are shown for species with a significant overall treatment effect on at least one survey date. Treatment codes: HOM = all patches fertilized at same time; HET = only some patches fertilized; HET50 and HET25 = 50% or 25% of patches fertilized at each feeding
SpeciesSurvey
Early 1998Late 1998
Overall treatment effectContrast: HOM vs. HETContrast: HET50 vs. HET25Overall treatment effectContrast: HOM vs. HETContrast: HET50 vs. HET25
A. stoloniferaNSNSNS< 0.05< 0.05NS
A. elatius< 0.02NS< 0.005< 0.000< 0.01< 0.001
D. glomerata< 0.01NS< 0.001< 0.001< 0.001< 0.01
F. ovina< 0.02NS< 0.005< 0.001< 0.001< 0.02
P. lanceolataNSNSNS< 0.05< 0.02NS
S. oleraceus< 0.05< 0.05NSNSNSNS
Rare species< 0.005< 0.05< 0.001< 0.05< 0.05NS
Figure 7.

Mean percentage cover (± SE) at three survey dates for species for which there was a significant or notable overall treatment effect (Table 7). Dark grey = early 1998, light grey = late 1998, open bars = early 1999. Results of planned contrasts are shown in Table 7 for significant differences between means and groups of means. See Methods for details of treatments. Treatment codes: HOM = all locations fertilized at same time; HET50P and HET50U = 50% of locations fertilized at each feeding either at the same (HET50P = predictable) or different (HET50U = unpredictable) locations; HET25P and HET25U = as above, with 25% of patches fertilized at each feeding.

Discussion

This experiment revealed highly significant effects of the pattern of nutrient supply on some facets of plant community structure, and more subtle effects on others. Community biomass and some aspects of community composition were significantly affected by the pattern of nutrient supply, even though we initiated each community from an identical inoculum of seeds, and supplied the same amount of nutrients to each treatment over the course of each growing season. However, there were relatively few significant effects on individual species or groups of species, given the number of tests performed, leaving impacts of pattern of nutrient supply at this level open to some doubt. There were also few effects on communities or their constituents caused by the patches of nutrient enrichment remaining the same or changing through time. Pattern of nutrient supply had no effect on the species richness or diversity of the above-ground vegetation.

community biomass and below-ground : above-ground biomass ratio

As predicted, productivity and below-ground : above-ground biomass ratio increased with patch density from HOM (100% enriched) to HET and from 50% to 25% enrichment. Significant increases in root biomass were accompanied by smaller increases in shoot biomass. This result supports our prediction that competing plants acquiring resources from a limited number of nutrient-rich patches invest more biomass in roots than plants in spatially homogeneous environments. The current experiment cannot, however, determine whether this is a direct result of mechanisms promoting underground competition, such as those associated with the Tragedy of the Commons model (Gersani et al. 2001), or of other possible mechanisms. Nevertheless, the production of more root biomass in the most heterogeneous treatments would be expected to reduce nutrient availability, and increase the intensity of below-ground competition in these treatments compared with the homogeneous treatment. We were surprised by the magnitude of the biomass response, with the lowest nutrient patch density leading to 71% more root biomass, and 44% more biomass overall, than the homogeneous treatment (Fig. 1), although the more limited increase in above-ground biomass suggests that fitness, measured as seed production, may not increase.

A partial explanation for the increased biomass observed in some treatments could be that although all treatments were given the same total quantity of nutrients, nutrient loss prior to plant uptake was greater in some treatments than others. At the start of the first year of the experiment, when plants were beginning growth, nutrients may have been lost more readily from the homogeneous treatment in which all of the nutrients were supplied at the outset, than from the heterogeneous treatments, in which a large proportion of the nutrients were provided after more extensive root systems had been produced. However, the effects of treatment on biomass (i.e. significant increases in both root biomass and in below-ground : above-ground biomass ratio in heterogeneous treatments) would not both be predicted on the basis of greater nutrient availability in these treatments.

species composition

A barrier to further interpretation of the current experiment is that roots of different species could not be distinguished. All the data on species composition, species abundance, species richness and diversity are based on above-ground parts, whereas the largest effects on biomass seen in this experiment were below-ground. Unfortunately, our experimental design, with a large number of replicates per block (leading to a low number of degrees of freedom associated with the block–treatment interaction term) combined with strong block × treatment interactions, also reduced our power to detect differences that a more powerful design might have identified. Nevertheless, we were able to demonstrate differences in above-ground community composition between treatments, with overall composition being strongly dependent on the pattern of nutrient supply (Fig. 3). Some individual species also showed significant responses to treatment, and the strength of their response to the contrast between homogeneous and heterogeneous treatments was clear. There were examples of both dominant (Dactylis glomerata) and relatively subordinate (Galium verum) species increasing significantly in biomass with decreasing nutrient patch density, in parallel with increase in overall biomass along this gradient. In contrast, Rumex crispus decreased in abundance along this gradient, and the colonist species, taken as a group, showed a more complex response. Species responses will depend not only on their response to the abiotic environment, but also on their response to each other. For instance, whether R. crispus is responding negatively to the abundance of neighbouring species, or directly to the pattern of nutrient supply, cannot be determined from this experiment.

The fact that species responded differently to the treatments suggests that some interspecific competitive relationships between species were affected by treatment (Fransen et al. 2001; Novoplansky & Goldberg 2001). Although we sowed a mixture of 20 species, the communities were rapidly dominated by a small subset of these species, most of which were perennial grasses (Table 1). This loss of species strongly suggests that competition played an important role in community development, although we have no direct evidence about its relative intensity in the different treatments.

percentage cover data

The percentage cover data from non-destructive surveys suggests that differences between communities in the five treatments arose very early in the experiment but that the effects of treatment on individual species continued to change throughout the experiment (Fig. 7), and that some strong early effects were lost over the course of the study. This is not surprising in an experiment in which nutrients were supplied heterogeneously in both time and space. At the time of the first survey in 1998, all HET treatments had received half their annual nutrient supply, yet significantly higher percentage cover values were recorded in the HET25 treatments than in the HET50 treatments. By the time of the later 1998 survey all treatments had received their full quota of nutrients for the year, and the lowest percentage cover values were recorded in the HOM treatment. Although percentage cover is a relatively imprecise measure of biomass, these data from the non-destructive harvests demonstrate clear patterns in species abundance arising as a result of the pattern of delivery of nutrients (and perhaps as a result of the capacity of plants to acquire nutrients when they were supplied, see above), and the dynamic nature of community development.

species richness and diversity

Our hypothesis concerning species richness and diversity was not correct, at least for above-ground vegetation. There was no significant effect of treatment on species richness at any survey (Fig. 2), and no effect on the Shannon indices of diversity and evenness at final harvest. Species richness showed a large and similar reduction in all treatments between the first and second years of the experiment. Given the differences in biomass and community composition (Fig. 3) between the treatments, it is perhaps surprising that there was no impact on the number of species they supported, or on diversity. These results add to the growing evidence showing that the relationship between diversity and heterogeneity need not be positive (Collins & Wein 1998a; Wilson 2000).

number of individual plants

The numbers of individuals of several species tended to be higher in heterogeneous than homogeneous treatments (Fig. 6). We had predicted that the environments with the fewest nutrient-rich patches (HET25 treatments) would generate the most intense competition, but this result suggests rather that establishment of some species at the start of the experiment was easier in the heterogeneous communities. However, fewer individuals of colonist species established in the HET25 treatments than in the HET50 and HOM treatments, suggesting that later establishment was more difficult at some nutrient patch densities than at others. Casper & Cahill (1996) and Day et al. (2003b) have shown that populations growing in nutritionally heterogeneous environments suffer lower overall mortality rates than populations grown with the same amount of nutrients supplied homogeneously. Day et al. (2003b) suggest that areas of low resource availability in patchy environments may provide refuges from strong competition because they support both low root biomass, and low above-ground biomass. Thus, more plants can survive in such patches, even though their size may be small. This could explain the greater number of plants surviving in the heterogeneous treatments in the current experiment.

predictions 1–3

The pattern of nutrient delivery had large impacts on community root biomass, community below-ground : above-ground biomass ratios and on the numbers and biomass of several individual species (Prediction 1). However, for the most part these effects seem to have been community-wide responses to pattern of nutrient delivery. Evidence for treatment effects on the relative abundance of different species in the community was less clear (Prediction 2). In general, there was higher biomass of species in the HET25 treatments than in the HET50 treatments. It is possible that the application of nutrients over a longer time period (as in the HET25 treatments compared with HOM) permits the acquisition of more nutrients and the accumulation of more biomass by species. An exception was Rumex crispus, which showed a significant decline in abundance from the HET50 to HET25 treatments.

Our results do not support the hypothesis that spatial predictability of nutrient supply is an important determinant of plant community structure when total nutrient supply is held constant (Prediction 3). This could be because those species that persisted in the developing communities were flexible in their ability to acquire nutrients under spatially unpredictable patterns of supply. For example, species characteristic of habitats in which resource supply is spatially and temporally infrequent and unpredictable are predicted to display efficient and localized physiological plasticity allowing rapid response to the appearance of nutrient patches (Crick & Grime 1987; Jackson et al. 1990). Few species responded to predictability of enrichment and only colonists showed a significant effect (Fig. 4c). At the final harvest their biomass was four times greater in the HET50P treatment than in the HET50U treatment. Again, the explanation for this could be that there were persistent refuges from strong competition in the predictable treatments in patches of substrate that were never enriched. In contrast, enriched sites where root competition was intense would have eventually covered the entire surface of most of the boxes in the unpredictable treatments. The strong competition that would be expected throughout the boxes in these treatments would be more limiting for late-establishing colonist species than for species present from the start of the experiment.

In conclusion, this study demonstrates important effects of heterogeneity in nutrient supply on plant community structure. In particular, biomass, below-ground : above-ground biomass ratio, and also, to a more limited extent, species composition and the sizes of some species populations were affected by the pattern of nutrient delivery. Thus, just as for individual plants, competing plants and monoculture populations, heterogeneity in resource supply has important consequences for plant communities which merit further exploration.

Acknowledgements

We gratefully acknowledge the help of a large number of people who contributed to this project. Principal among them were Alison Burtenshaw, Patrick Fitzsimons, Sophie Homewood, Cliff Hughes, Audra Hurst, Rhian Langford, Jessie Leamy, Steve McAuliffe, Duncan Ponter, Arie Ramp, Katie Ramp-Vasey, Morag Schuaib, Martyn Stenning, Tatiana White and Annie Wright. The manuscript has been improved by comments from two anonymous referees, Tong Davy, Steve Coad, Zoltan Dienes and Sue Hartley. This research was funded by NERC grant (GR3/11069) awarded to EAJ and MJH.

Ancillary