The effects of spatially heterogeneous nutrient supply on yield, intensity of competition and root placement patterns in Briza media and Festuca ovina

Authors


E.A. John. E-mail: e.a.john@sussex.ac.uk

Summary

  • 1An experiment was carried out to determine whether the spatial pattern of soil nutrient distribution influences plant growth and root placement, and the intensity of competition between plants. Our hypotheses were that (i) growth would be greater for plants in heterogeneous than homogeneous conditions (ii) competitive intensity would be greater when nutrients were confined to a small proportion of the substrate volume than when uniformly distributed, and (iii) plants would selectively place roots in nutrient-rich patches in heterogeneous conditions.
  • 2We grew plants of Briza media and Festuca ovina alone, or with an intra or interspecific competitor, in pots in which the same total supply of nutrients was distributed either homogeneously or heterogeneously. Nutrients in the heterogeneous treatment were confined to three soil cores occupying only 6·5% of the substrate volume.
  • 3Pattern of nutrient supply did not affect yield of plants grown alone. However, the reduction in target plant yield caused by competition was much greater in heterogeneous than homogeneous conditions (i.e. competitive intensity was greater in heterogeneous conditions).
  • 4Plants growing in heterogeneous conditions placed roots preferentially in the nutrient-rich substrate patches. This may have been the cause of greater competitive intensity in heterogeneous conditions. It is speculated that soil-based heterogeneity might exert an important influence on community structure and composition through its effects on competition.

Introduction

All natural soils exhibit heterogeneity in nutrient supply (Jackson & Caldwell 1989, 1993; Caldwell et al. 1991; Caldwell & Pearcy 1994; Richard et al. 2000). When the scale of heterogeneity is small enough for individual plants to experience patches of different quality, localised physiological or morphological responses may be exhibited. Some species utilise morphological specialization and selective growth of roots in nutrient-rich patches to forage efficiently for nutrients (Drew 1975; Crick & Grime 1987; Jackson & Caldwell 1989; Hutchings & de Kroon 1994; Robinson 1994; Alpert & Stuefer 1997; Einsmann et al. 1999). Gersani et al. (1998) and Wijesinghe & Hutchings (1999) have reported that the relative quality of different soil patches can determine the extent to which nutrient-rich patches are selected for root placement. Morphological responses to heterogeneity differ between species (Crick & Grime 1987; Campbell & Grime 1989; Schottelndreier & Falkengren-Grerup 1999; Wijesinghe et al. 2001) and are nutrient-specific (Drew 1975; Zhang & Forde 1998; Robinson 1994; Schottelndreier & Falkengren-Grerup 1999). Physiological responses to heterogeneity, such as higher uptake rates in nutrient-rich patches, and characteristics of the substrate such as increased ion availability at higher concentrations, may all lead to greater nutrient uptake in heterogeneous environments than in homogeneous environments with the same overall nutrient supply (Jackson & Caldwell 1996). Such ‘efficiency gains’ may explain why some species produce more biomass in heterogeneous conditions than in homogeneous conditions providing the same supply of nutrients (Birch & Hutchings 1994; Einsmann et al. 1999; Wijesinghe et al. 2001).

Despite the growing evidence of effects of nutrient heterogeneity on isolated plants, there is little evidence that it affects competition (Schwinning & Weiner 1998; Cahill & Casper 1999, 2000), although recent research suggests that competitive hierarchies can be altered by resource heterogeneity (Fransen et al. 2001; Novoplansky & Goldberg 2001). It might be predicted, however, that localisation of nutrients within patches would increase the intensity of competition, as nutrient-rich patches can contain roots of several plants (Casper et al. 2000). Robinson et al. (1999) showed that nitrogen capture by competing plants in nutritionally heterogeneous conditions was strongly correlated with their rates of root proliferation in nitrogen-rich patches; species with faster root proliferation gained a competition advantage. This also suggests that nutrient heterogeneity may affect intensity of competition. We therefore conducted an experimental study to examine whether the spatial pattern of nutrient supply influences plant root placement and the intensity of competition. Plants were grown in either patchy or homogeneous conditions providing the same total nutrient supply, and either alone or in competition. The hypotheses tested, and the predictions associated with them, were as follows:

Hypothesis 1: Total yield of plants grown alone and in competition will be greater when nutrient distribution is heterogeneous than when it is homogeneous. It was predicted that selective root placement and root proliferation in nutrient-rich patches in heterogeneous conditions would allow greater nutrient acquisition than in homogeneous conditions, resulting in greater yield in heterogeneous conditions.

Hypothesis 2: The intensity of competition between plants would be greater in heterogeneous conditions than in homogeneous conditions providing the same total supply of nutrients. Specifically, it was predicted that competing plants would have lower yields than plants grown alone, and that yield reduction due to competition would be greater in heterogeneous conditions.

Hypothesis 3: Plants growing in heterogeneous conditions will place more root biomass in nutrient-rich patches than plants growing in homogeneous conditions will place in equivalent locations, i.e. plants in heterogeneous conditions will forage for nutrients through preferential root placement in nutrient-rich patches (Hutchings & de Kroon 1994).

Methods

The species used in the experiment were Briza media and Festuca ovina (both Poaceae). These tuft-forming perennial grasses both grow and compete in nutrient-poor grassland communities on calcareous soils (Tansley 1939). Both species have low growth rates and similar sizes, and are stress-tolerators (sensuGrime & Hunt 1975). They have similar morphological and growth responses to nutrient supply (Day 2001).

The experiment was carried out in a glasshouse maintained at 20 ± 5 °C under a 16-h photoperiod. Natural daylight was supplemented by light from Osram SON-T 400 W lamps. The growing medium was prewashed potting sand (Fargro Ltd, Littlehampton, UK), with nutrients in the form of the controlled release fertiliser Osmocote Mini Plus (Scotts Professional Limited, Nottingham, UK) supplied either uniformly or in patches. Daily rates of nutrient release are given in Day (2001; see also Huett 1997). Total nutrient supply was the same in all treatments. All pots contained 275 mg (0·375 mg l−1) of fertiliser, equating to 91·5 mg per nutrient-rich patch (6 mg l−1) in heterogeneous treatments and 5·9 mg per equivalent patch of substrate in homogeneous treatments. Nutrient-rich patches were created in heterogeneous treatments by removing three vertical cylindrical cores (1·5 cm in diameter, 6·5 cm deep), located equidistantly on the pot diameter, and replacing them with a sand/fertiliser mix. The higher nutrient status of these three cores compared with the background substrate was shown by algal growth only being apparent on the surfaces of the three cores. Two virtual patches with the same dimensions were also created, from which cores of sand were removed, and replaced with sand as a procedural control (Fig. 1). Nutrients were distributed evenly throughout the substrate in the homogeneous treatment. Cores were also removed and replaced in this treatment as a control. The five cores were identified as follows in all treatments (Fig. 1): (a) Target's patch; the patch close to the target plant (see below) and most distant from the competitor (when present) (b) Competitor's patch; the patch close to the competitor (when present) and most distant from the target plant (c) Shared patch; the patch equidistant from, and directly between, the target plant and competitor (when present) (d) Target's virtual patch; the patch close to the target plant and at the same distance (1·75 cm) from this plant as the target's patch and the shared patch (e) Competitor's virtual patch; the patch close to the competitor (when present) and at the same distance (1·75 cm) from this plant as the competitor's patch and the shared patch. Nutrient-enriched cores (a–c) in the heterogeneous treatment made up 6·5% of the substrate by volume.

Figure 1.

Schematic diagram of the experimental set-up for treatments with competitor plants. In treatments with only target plants no competitor was planted. See text for details.

Seeds of both species (from Emorsgate Seeds Ltd, Kings Lynn, UK) were germinated on vermiculite. After 14 days, seedlings were selected for uniformity of root and shoot size and transplanted into 12·5 cm diameter pots placed on a wire mesh bench so that excess water and nutrients could escape without causing contamination between pots. Pots were watered as required. Either one or two seedlings were planted per pot. Plants grown alone (designated target plants) were located directly between the target's patch and the shared patch, and equidistant from these two patches and the target's virtual patch. When two plants were grown in a pot, the target plant was placed as above while the other (the competitor) was located directly between the competitor's patch and the shared patch, and equidistant from these two patches and the competitor's virtual patch. Each species was grown alone as the target species, and in competition, with either an intraspecific or interspecific neighbour, under both homogeneous and heterogeneous conditions. Thus there were 12 treatments in all: two species as target, planted in three neighbourhood arrangements (alone, intraspecific competitor, interspecific competitor), under two nutrient arrangements (homogeneous or heterogeneous). Each treatment was replicated 10 times. Pots were laid out in a randomised block design with one replicate of each treatment in each block.

Plants were harvested when flower spikes were first observed, after 67 days. By this time, in all 12 treatments, between one and three replicates had suffered the death of either one or both plants. Above- and below-ground parts were harvested separately. Above-ground parts of each plant were placed in separate labelled bags, oven-dried at 80 °C to constant weight, and weighed. To harvest below-ground parts, a specially designed tool was used to separate all five patches and the remaining volume of substrate. Roots were extracted from each fraction, washed, placed in separate labelled bags, oven-dried at 80 °C to constant weight, and weighed. In a few cases patches collapsed during extraction. Thus in some treatments the number of replicates available for the analysis of root biomass in different patches was slightly less than that for the analysis of above-ground mass and total root mass.

data analysis

Data were analysed using SPSS version 8·0 for Windows (Generalised linear model procedure SPSS Inc., 1989–97). Data were normally distributed and homoscedastic. There were no effects of blocks or of block–treatment interactions, so blocks were dropped from all analyses.

Hypothesis 1: Yield and Biomass Allocation

One-way multivariate analyses of variance (manova) were carried out to determine the effects of treatment (pattern of nutrient supply and neighbourhood) on yield. Separate analyses were conducted for each species as target, with treatment as the independent variable, and above- and below-ground biomass as dependent variables. This was followed by the corresponding univariate analyses of variance (anova). A univariate anova was used to analyse the effect of treatment on total biomass. Significant differences between means were determined using Bonferroni multiple-means comparison tests at P < 0·05.

Hypothesis 2: Competition

To investigate the effect of treatment on the intensity of competition in heterogeneous and homogeneous environments, mean competitive intensity (CI) was calculated for the target plants. As an index of absolute competitive intensity, we used the mean reduction in yield in the presence of a competitor compared with mean yield in the absence of a competitor. A formula of Kadmon (1995) was adapted to calculate CI (in g) experienced by plants grown with a competitor:

CI experienced by the target species when grown with an intraspecific competitor:

CI = [2Yi,sp1– Yc,sp1]/2((equation 1))

CI experienced by both species when grown with an interspecific competitor:

CI = [(Yi,sp1 + Yi,sp2) – Yc,sp1]/2,((equation 2))

where Y = yield, i denotes plants grown alone, c denotes plants grown in competition (and therefore always two plants), sp1 is the target plant species and sp2 is the competitor species. Biomass data from each replicate target plant of a given species, in each species combination, were matched by nutrient distribution with the mean biomass for plants of the same species grown alone. For the interspecific competition treatment, CI was calculated for both species growing together, as it was impossible to separate their below-ground parts. However, CI was calculated separately for cases of intraspecific competition in which B. media and F. ovina were the target plants.

If plants grown together are unaffected by competition, CI is zero. Positive CI values indicate that plants are competing, and negative values indicate that they benefit from each other's presence. To determine whether CI differed significantly from zero, one-sample t-tests were carried out. The effects of treatment (pattern of nutrient supply and neighbour identity) on CI were investigated using one-way anova with treatment as independent variable and CI as dependent variable. Significant differences between mean values were determined using Bonferroni multiple-means comparison tests at P < 0·05.

Hypothesis 3: Foraging

To analyse the effect of treatment and patch location on biomass of roots in the five substrate patches, a one-way split-plot analysis of variance (spanova, von Ende 1999) was carried out with pot as the split treatment, treatment (pattern of nutrient supply and neighbourhood) as the between-treatment factor and patch location as the within-treatment factor. The dependent variable was root biomass in each of the five patches. When within-treatment effects were significant, anovas were carried out to determine the source of variation, with patch as the independent variable and root biomass as the dependent variable. When F-values were significant, Bonferroni multiple-means comparisons (at P < 0·05) were used to determine which means differed significantly.

Results

effect of treatment on yield and biomass allocation

Multivariate and univariate analyses of variance revealed that, for both species as target, above-ground biomass, below-ground biomass and total biomass were significantly affected by treatment (Table 1). Pattern of nutrient supply did not significantly affect any measure of yield when either species was grown alone (Figs 2a,b). In the absence of competitive or beneficial effects of neighbours, the yield of two plants of the same species grown together would be twice that of a plant growing alone. Under both patterns of nutrient supply, mean yields in treatments involving intraspecific neighbours were usually considerably less than twice those of plants grown alone, suggesting that competition had occurred. In heterogeneous conditions, mean total biomass, and above- and below-ground biomass for both B. media and F. ovina as targets, grown with neighbours, was greater, but not significantly so, than when they were grown alone. In homogeneous conditions, the mean total biomass of B. media as target grown with a B. media neighbour was close to twice that of B. media grown alone (Fig. 2a), suggesting that competition was weak, but with F. ovina as neighbour, the combined yield of both plants was not significantly greater than that of a single B. media plant, indicating strong competition (Fig. 2a). When F. ovina was the target species in homogeneous conditions, the combined yield of two F. ovina plants was not significantly greater than that of one F. ovina plant grown alone (Fig. 2b), indicating that these plants competed. However, when F. ovina grew with B. media, their mean combined yield was over twice that of F. ovina grown alone, indicating a lack of competition (Fig. 2b).

Table 1.  The effect of treatment on above- and below-ground biomass of Briza media (a) and Festuca ovina (b). The table shows the results of multivariate anovas and associated univariate anovas with treatment as independent variable and above- and below-ground biomass as dependent variables. The results of independent univariate anovas investigating the effect of treatment on total biomass are also shown. Values used in the analysis were weights of either single plants (in treatments without competition) or summed weights of two plants (in treatments with competition). d.f. = degrees of freedom, SS = sum of squares, MS = mean square. See Fig. 2 for means and standard errors
(a)                    B. media as the target species
SourceWilks’ Lambdad.f.  F-valueP
Treatment0·5710,84  2·750·006
SourceVariabled.f.SSMSF-valueP
TreatmentAbove-ground biomass 5 3·860·774·11    0·004
ErrorAbove-ground biomass43 8·070·19  
TreatmentBelow-ground biomass 5 8·041·613·31    0·013
ErrorBelow-ground biomass4320·910·49  
TreatmentTotal biomass 521·74·343·87    0·006
ErrorTotal biomass4348·161·12  
(b)                    F. ovina as the target species
SourceWilks’ Lambdad.f.  F-valueP
Treatment0·4510,76  3·70< 0·001
SourceVariabled.f.SSMSF-valueP
TreatmentAbove-ground biomass 5 5·481·107·14< 0·001
ErrorAbove-ground biomass39 5·990·15  
TreatmentBelow-ground biomass 510·902·185·61< 0·001
ErrorBelow-ground biomass3915·160·39  
TreatmentTotal biomass 531·156·236·87< 0·001
ErrorTotal biomass3935·490·91  
Figure 2.

The effects of treatment on above-ground biomass, below-ground biomass and total biomass of Briza media (a) and Festuca ovina (b) as target plants. Graphs show mean values for either single plants (Alone = treatment without competition) or summed weights for two plants (Intra and Inter = treatments with intra- and interspecific competition, respectively). All treatments provided the same total quantity of nutrients. Bars show 1 SE of total means (upward bars) and component means (downward bars). Letters indicate significant differences between treatments for mean total biomass (above error bars) and component biomass (below error bars) at the P < 0·05 level, following Bonferroni multiple-means comparison tests. See Table 1 for associated analyses.

effect of treatment on competitive intensity

Competitive intensities were significantly affected by pattern of nutrient supply in treatments where B. media and F. ovina were target plants (Table 2, Fig. 3). The estimated biomass of the target plant in heterogeneous treatments was significantly affected by competition. For B. media as target, CI only differed significantly from zero in heterogeneous treatments (intraspecific neighbour; t= 5·526, d.f. = 6, P= 0·001: interspecific neighbour; t= 2·912, d.f. = 7, P= 0·023, Fig. 3a). For F. ovina as target, CI differed from zero in homogeneous treatments with an intraspecific neighbour (t = 4·472, d.f. = 6, P= 0·004, Fig. 3b) and in heterogeneous treatments with both intra- and interspecific neighbours (intraspecific; t= 2·641, d.f. = 7, P= 0·033: interspecific; t= 3·260, d.f. = 7, P= 0·014). However, there was no evidence that F. ovina competed with B. media in the homogeneous treatment, because CI did not differ significantly from zero. In summary, B. media competed with neighbours in heterogeneous conditions, but there was no evidence of competition in homogeneous conditions. F. ovina competed with neighbours in heterogeneous conditions. In homogeneous conditions F. ovina competed with intraspecific neighbours but not with interspecific neighbours.

Table 2.  The effect of treatment on competitive intensity (CI) with B. media (a) and F. ovina (b) as target plant. Results of univariate analyses of variance. d.f. = degrees of freedom, MS = mean square. See Fig. 3 for means and standard errors
(a)            B. media as the target plant
Dependent VariableSourced.f.MSF-valueP
CITreatment 30·6633·5840·027
 Error270·185  
(b)            F. ovina as the target plant
Dependent VariableSourced.f.MSF-valueP
CITreatment 30·7393·3400·035
 Error260·221  
Figure 3.

Mean (+ 1 SE) competitive intensity in treatments with B. media (a) and F. ovina (b) as target plant. Nutrients were either spatially heterogeneous (dark columns) or spatially homogeneous (light columns). Letters indicate significant differences in mean CI values between treatments at the P < 0·05 level, following Bonferroni multiple-means comparison tests. See Table 2 for associated analyses.

effect of treatment on foraging

For B. media and F. ovina, treatment (nutrient distribution and neighbour identity) and patch location significantly affected root biomass recovered from patches, both within and between treatments. Significantly more root biomass was produced in nutrient-rich patches in all heterogeneous treatments than in equivalent patches in the homogeneous treatments. In contrast, when competitors were present, there was more root mass in the virtual patches in the homogeneous treatment than in the heterogeneous treatment (Tables 3 and 4, Figs 4 and 5).

Table 3.  The effects of treatment (neighbourhood and nutrient distribution) (T) and location within treatment (L) on biomass of roots in patches when B. media (a) and F. ovina (b) were the target plants. Results of two-way split-plot analyses of variance comprising multivariate analyses of variance on location and location × treatment [i], between-treatments analyses of variance [ii], and within-treatments analyses of variance on location and location × treatment [iii]. In treatments where competitors were present the biomass values were derived from roots of two plants. All treatments provided the same total quantity of nutrients. d.f. = degrees of freedom, SS = sum of squares and MS = mean square. See Figs 4 and 5 for means and standard errors
(a)              Analyses for treatments with B. media as the target species
Multivariate Tests
[i] SourceWilks’ Lambdad.f.F-valueP
L 0·1874,35 37·931< 0·001
L × T 0·05320,117  8·303< 0·001
Between-Treatment Tests
[ii] Sourced.f.SSMSF-valueP
T  50·0213 0·0043    9·118< 0·001
Error 380·0178 0·0005  
Within-Treatments Tests
[iii] Sourced.f.SSMSF-valueP
L  40·0189 0·0047   62·749< 0·001
L × T 200·0258 0·0013   17·070< 0·001
Error1520·0115 0·00007  
(b)             Analyses for treatments with F. ovina as the target species
Multivariate Tests
[i] SourceWilks’ Lambdad.f. F-valueP
L  0·4004,32 12·025< 0·001
L × T  0·18220,107  3·601< 0·001
Between-Treatment Tests
[ii] Sourced.f.SSMSF-valueP
T  50·0222 0·0044    7·230< 0·001
Error 350·0214 0·0006  
Within-Treatments Tests
[iii] Sourced.f.SSMSF-valueP
L  40·0109 0·0027   32·110< 0·001
L × T 200·0153 0·0007    8·955< 0·001
Error1400·0119 0·0001  
Table 4.  Within-treatment analyses of the effects of patch location on root biomass in patches when B. media (a) and F. ovina (b) were the target species. Twelve independent one-way univariate analyses of variance are reported, with patch location as the fixed-factor and root biomass in the target's patch, shared patch, competitor's patch, target's virtual patch and competitor's virtual patch as the dependent variables. In treatments where competitors were present the biomass values were derived from roots of two plants. All treatments provided the same total quantity of nutrients. d.f. = degrees of freedom. See Figs 4 and 5 for means and standard errors
(a)              Analyses for treatments with B. media as the target species
TreatmentNutrient DistributionSpecies Compositiond.f.F-valueP
1HeterogeneousAlone4,35 27·341< 0·001
2HeterogeneousIntraspecific4,25107·448< 0·001
3HeterogeneousInterspecific4,30  9·656< 0·001
4HomogeneousAlone4,35  0·839    0·510
5HomogeneousIntraspecific4,35  0·281    0·888
6HomogeneousInterspecific4,30  0·377    0·823
(b)              Analyses for treatments with F. ovina as the target species
TreatmentNutrient DistributionSpecies Compositiond.f.F-valueP
7HeterogeneousAlone4,25  6·5780·001
8HeterogeneousIntraspecific4,30 20·4550·001
9HeterogeneousInterspecific4,30  2·7870·050
10HomogeneousAlone4,35  2·168    0·093
11HomogeneousIntraspecific4,25  0·690    0·605
12HomogeneousInterspecific4,30  0·267    0·897
Figure 4.

The effects of treatment on the mean (+ 1 SE) biomass of roots recovered from different patches with B. media as the target plant. Graphs show root biomass recovered from the target's patch (a), shared patch (b), competitor's patch (c), target's virtual patch (d) and competitor's virtual patch (e). In treatments where a competitor was present the values shown included roots of two plants. Letters above the error bars indicate significant differences between treatments at the P < 0·05 level following a Bonferroni multiple-means comparison test. See Tables 3 and 4 for associated multivariate and univariate analyses.

Figure 5.

The effects of treatment on the mean (+ 1 SE) biomass of roots recovered from different patches with F. ovina as the target plant. Graphs show root biomass recovered from the target's patch (a), shared patch (b), competitor's patch (c), target's virtual patch (d) and competitor's virtual patch (e). In treatments where a competitor was present the values shown included roots of two plants. Letters above the error bars indicate significant differences between treatments at the P < 0·05 level following a Bonferroni multiple-means comparison test. See Tables 3 and 4 for associated multivariate and univariate analyses.

In homogeneous treatments with B. media as target, root biomass per patch was significantly affected by the number and identity of plants present (Fig. 4). When an intraspecific neighbour was present, root biomass in patches was at least double that in the treatments where plants were grown alone. Far less increase was seen when an interspecific neighbour was present, particularly in the nutrient-rich patches. In heterogeneous conditions the amount of root biomass per patch was generally not affected by the number of plants in the treatment, but an exception to this was the competitor's patch, from which twice as much biomass was recovered when an intraspecific neighbour was present as when B. media grew alone or with an interspecific neighbour (Fig. 4c).

When F. ovina was the target species, the root biomass in the five patches in the homogeneous treatment depended both on plant number and on neighbour identity. With an intraspecific neighbour, the biomass in all of the patches was not significantly different from that when plants were grown alone. With an interspecific neighbour, the root mass in each of these patches was at least three times greater than when plants were grown alone (Fig. 5). In heterogeneous treatments there were no treatment-specific significant differences in the biomass recovered from particular patches, with the exception of the competitor's patch, where more biomass was present in the treatments with an intraspecific competitor than in the treatment with F. ovina growing alone (Fig. 5c).

For both species in heterogeneous treatments, patch root biomass depended on patch location (Table 4). The root biomass recovered from both the target's patch and the shared patch was at least three times that in either virtual patch (Figs 4 and 5), demonstrating that both species responded to localised nutrient enrichment. However, when nutrients were homogeneously distributed, there was no effect of patch location on root biomass (Table 4, Figs 4 and 5), supporting Hypothesis 3.

When B. media was the target species, the three nutrient-rich patches contained similar amounts of root in the treatment with an interspecific neighbour, but with an intraspecific neighbour the root mass in the shared patch was significantly less than in either the target's or competitor's patches (compare Fig. 4a,b,c). When F. ovina was the target species, all three nutrient-rich patches contained similar root masses in treatments with both intraspecific and interspecific neighbours (Fig. 5).

Discussion

The major results relating to our hypotheses are as follows. (1) The pattern of nutrient supply did not affect yield in treatments with the same species composition (Fig. 2a,b). Thus, Hypothesis 1 was not supported. (2) In treatments where B. media competed intraspecifically and where B. media and F. ovina competed together, the intensity of competition was significantly greater in heterogeneous conditions than in homogeneous conditions with the same nutrient supply, strongly supporting Hypothesis 2 (Fig. 3). (3) Both species responded to local soil nutrient enrichment, as shown by differences in root biomass in substrate patches of different quality (Figs 4 and 5), providing strong support for Hypothesis 3. The implications of these results are discussed below.

hypothesis 1

Pattern of nutrient supply did not affect total yield, or above- or below-ground yield (Fig. 2a,b) of B. media or F. ovina grown alone. These results contrast with those of several recent studies that have shown significant yield increases when a fixed supply of nutrients is provided heterogeneously to single plants (e.g. Birch & Hutchings 1994; Alpert & Stuefer 1997; Wijesinghe & Hutchings 1997, 1999). These studies were on clonal plants with many rooting sites. Clonal plants growing in heterogeneous conditions benefit from division of labour, with ramets in different quality habitat patches specialising to acquire locally abundant resources. Studies using plants with a single rooting site show more variation in whether or not species respond to heterogeneity with a change in overall biomass (Einsmann et al. 1999; Wijesinghe et al. 2001; Bliss et al. 2002). For instance, Wijesinghe et al. (2001) found that yield of two out of six nonclonal herbaceous species was altered by pattern of nutrient supply, and Einsmann et al. (1999) reported that four out of 10 herbaceous and woody species had higher yields under heterogeneous conditions. However, in a different experiment on some of the same species used by Einsmann et al. (1999), Bliss et al. (2002) found that none showed yield changes in response to heterogeneity. Such contrasting results suggest that whether or not a species responds positively to heterogeneity is probably context-specific (Wijesinghe et al. 2001; Bliss et al. 2002). Although our hypothesis was based on the assumption that more efficient nutrient acquisition from patchy substrate (Jackson & Caldwell 1996) would benefit growth, it is perhaps more surprising that this extreme test of the impact of heterogeneity, in which nutrients were confined to only 6·5% of the substrate volume, did not reduce growth. That isolated plants could achieve the same yield in heterogeneous and homogeneous conditions implies very efficient extraction of resources from a very small volume of substrate.

hypothesis 2

The prediction that competition intensity would be greater under heterogeneous nutrient supply (Robinson et al. 1999) was upheld for B. media, and for F. ovina grown with interspecific neighbours (Fig. 3). We suggest that, as the nutrient-rich patches accounted for such a small proportion of the substrate volume in the heterogeneous treatments, this led to more intense competition for the same quantity of nutrients than in the homogeneous treatments. Robinson et al. (1999) and Hodge et al. (1999) have proposed that root proliferation in nutrient patches may have evolved as a competition mechanism, because nutrients can be more effectively depleted from discrete patches. If greater intensity of competition in heterogeneous conditions is a general phenomenon, it would have important implications for the coexistence and relative abundance of species in communities, and for the value of studies of community structure under homogeneous conditions. The impact of soil heterogeneity on the intensity of competition is also likely to depend on the form of that heterogeneity. For instance, plants may compete more strongly for nutrients from a single shared patch than for the same quantity of nutrients distributed across several patches. As far as we are aware the impact of nutrient heterogeneity on intensity of competition has not yet been studied in field situations, and thus critical experiments to test such predictions are still awaited. It is possible that plants experience more diffuse competition under natural conditions than in experiments, and that this would moderate the intensification of competition demonstrated here in heterogeneous environments. Heterogeneity in nutrient supply can have significant effects on the interactions between plants growing in natural communities, as shown, for example by the fact that the composition of field-grown plant communities can be strongly affected by the pattern of nutrient supply (Wijesinghe et al. submitted). However, the mechanisms involved await elucidation. Further research on this area is warranted.

hypothesis 3

B. media and F. ovina produced more root biomass in nutrient-rich patches than in ‘virtual’ patches of the same size. Thus, the spatial distribution of roots was clearly affected by nutrient availability. Selective placement of roots in nutrient-rich soil patches is a well-known response to soil heterogeneity, and indicates that plants assess environmental quality at scales smaller than that of their whole root system, and invest root biomass appropriately for efficient resource acquisition (Gersani et al. 1998; Wijesinghe & Hutchings 1999). Despite this, the ability to place roots preferentially in nutrient-rich patches did not increase overall biomass. This may have been because both the pots and the nutrient-rich cores in the heterogeneous treatments were small enough for the plants to fully exploit the available nutrients before harvest, regardless of their pattern of supply (Hutchings et al. 2003).

The results of this study demonstrate that the ability of plant roots to respond to heterogeneity in nutrient supply can have a significant impact on the intensity of competition. The cause may be the tendency of plant roots to aggregate in nutrient-rich patches. The responses of plant species to soil nutrient heterogeneity are complex, but we suggest that heterogeneity in nutrient supply, acting through its effect on competition, is likely to be an important factor controlling the structure of plant communities.

Acknowledgements

This research was carried out while K.J.D. was in receipt of a University of Sussex studentship. We acknowledge practical help from David Aplin during the experimental, analytical and writing phases of this experiment, and helpful comments from Sue Hartley, and an anonymous referee on drafts of the manuscript.

Ancillary