The effects of spatial pattern of nutrient supply on yield, structure and mortality in plant populations

Authors


M. J. Hutchings (tel. +44 1273 872761, fax +44 1273 678433, e-mail M.J.Hutchings@sussex.ac.uk).

Summary

  • 1Early in growth, plant populations subjected to heterogeneous conditions can achieve significantly higher yields than populations grown in homogeneous conditions that provide the same amount of nutrients. We hypothesized that this yield enhancement would be ephemeral, and that yields of populations grown under different patterns of nutrient supply will converge as time passes. We also predicted that plant size frequency distributions will continue to differ between populations in heterogeneous and homogeneous conditions, that spatial patterns of mortality will differ, and that greater mortality will occur in populations under heterogeneous conditions. An experiment was carried out on Cardamine hirsuta to test these hypotheses.
  • 2The experiment was harvested after 60 days, when flowering had begun in all populations. At this time, populations grown with the same nutrient supply but different patterns of supply did not differ in yield. Populations grown at higher nutrient levels had greater yields but suffered more mortality.
  • 3In heterogeneous conditions, plants in nutrient-rich locations were larger than plants in nutrient-poor locations, and larger than plants in identical positions in populations growing under homogeneous conditions. Plants in nutrient-rich patches in heterogeneous treatments suffered more mortality than plants in nutrient-poor patches. Contrary to prediction, total mortality at all levels of nutrient supply was significantly higher in populations in homogeneous conditions.
  • 4Coefficient of variation in shoot sizes was higher in populations at higher nutrient levels, but unaffected by pattern of nutrient supply. CV of root mass per unit of substrate was greater under heterogeneous than homogeneous conditions.
  • 5Frequency distributions of shoot size, and root mass per unit of substrate, were strikingly different between populations grown under different patterns of nutrient supply. The differences are interpreted as showing that under homogeneous conditions all parts of the habitat are heavily exploited, giving rise to uniformly high levels of interplant competition. In contrast, nutrient-poor patches in the heterogeneous treatments contain low root masses. Plants in these patches remain small, but have a high probability of survival. We propose that these differences in intensity of habitat occupation, and presumably in competition, explain the lower mortality in populations growing under heterogeneous conditions.

Introduction

Several studies have shown that heterogeneity in nutrient supply can increase the yield of individual plants compared with their growth under homogeneous conditions supplying the same quantity of nutrients (Birch & Hutchings 1994; Alpert & Stuefer 1997; Hutchings & Wijesinghe 1997; Wijesinghe & Hutchings 1997, 1999; Einsmann et al. 1999). This is achieved through responses such as selective use of more favourable habitat patches, and increased rooting in nutrient-rich patches (Drew 1975; Jackson & Caldwell 1989; Hutchings & de Kroon 1994; Robinson 1994; Wijesinghe & Hutchings 1997, 1999; Fransen et al. 1998; Einsmann et al. 1999; Robinson et al. 1999). Facelli & Facelli (2002) and Day et al. (2003a) have recently shown that yield of populations can also be enhanced by heterogeneity in nutrient supply. Day et al. (2003a) showed that, after 31 days of growth, Cardamine hirsuta populations growing under heterogeneous conditions achieved greater total yields, and displayed greater variation in yield both between individual plants and between patches of substrate, than populations in homogeneous conditions providing the same amount of nutrients. The proportion of the biomass located in nutrient-rich patches also increased significantly as patch size increased, and size of individual shoots was strongly dependent upon whether they were growing in nutrient-rich or nutrient-poor patches of substrate.

Heterogeneity in nutrient supply thus enhanced population yield and affected several other aspects of population performance, presumably because resources were acquired more efficiently than under homogeneous conditions during this early phase of growth. It was speculated, however, that this yield enhancement might be ephemeral. If populations had been grown for longer, their yields might have converged, because all had been given the same total supply of nutrients. At the same time it was predicted that, because roots proliferate strongly in nutrient-rich patches, these patches would become sites of intense competition, leading to heavy local mortality as growth continues. Schwinning & Weiner (1998) have hypothesized that below-ground competition will be asymmetric when nutrient distribution is heterogeneous, because plants with larger root systems would be more likely to encounter and dominate nutrient-rich patches. Recent empirical evidence (Day et al. 2003b) suggests that intensity of below-ground competition between plants is heightened by heterogeneity in nutrient supply. In contrast, nutrient-poor patches may be sites of less intense competition, where plants remain small but have a high probability of survival. Consequently, while we predict convergence in the yields of populations growing under different patterns of nutrient supply as growth continues, frequency distributions of plant sizes will continue to differ, being dependent on pattern of nutrient supply (see also Fransen et al. 2001). Spatial patterns of self-thinning, and overall mortality should also depend on pattern of nutrient supply. Thus, although heterogeneity in nutrient supply may not affect population yield in the long term, it would continue to have significant impacts on many other aspects of population structure.

This paper describes an experiment examining these predictions about plant population responses to pattern of nutrient supply. The effects of nutrient supply and heterogeneity in nutrient supply on populations of Cardamine hirsuta (Brassicaceae) were investigated. Population yield, shoot size inequality, variation in root biomass per unit of substrate and the spatial pattern of biomass distribution were measured in populations grown in uniform conditions and in heterogeneous conditions with two patch scales. These patterns of nutrient supply were crossed with three different levels of total nutrient supply. Mortality and its distribution was also examined. The hypotheses tested, and the predictions associated with them, were as follows.

Hypothesis 1. Yields of plant populations growing under homogeneous and heterogeneous conditions with the same total nutrient supply will not be significantly different when density-dependent self-thinning is occurring, but populations with higher nutrient supply will have achieved greater yields at the same stage of growth.

Hypothesis 2. Faster growth in populations with higher nutrient supply will produce greater variation in individual shoot size and root mass per unit of substrate. Because of differences in growth potential between nutrient-rich and nutrient-poor patches, variation in shoot size and root mass per unit of substrate will be more pronounced in populations growing under heterogeneous conditions than under homogeneous conditions at the same overall level of nutrient supply.

Hypothesis 3. Prior to self-thinning in populations, the spatial distribution of biomass depends on pattern of nutrient supply, and the proportion of biomass in nutrient-rich patches is greater at larger patch scales (Day et al. 2003a). When density-dependent self-thinning has begun, total population yield is predicted to be the same in all treatments with the same nutrient supply. We also predict that the proportions of biomass located in rich (and poor) patches will be independent of patch scale.

Hypothesis 4. Greater variation in shoot size and root mass per unit of substrate in populations in heterogeneous conditions will lead to more severe competition than in homogeneous conditions. This will ultimately result in a higher level of mortality in populations in heterogeneous conditions.

Materials and methods

Cardamine hirsuta (hairy bitter cress, Brassicaceae) is a small, usually annual herb with a basal rosette of numerous pinnate leaves arranged in a compact rosette (Grime et al. 1988). Flower stems can be 30 cm tall in nutrient-rich conditions (Stace 1997).

The experiment was carried out in a glasshouse maintained at a temperature of 20 ± 5 °C under a 16-h light : 8-h dark cycle. Natural daylight was supplemented by light from Osram VIALOX 400 W lamps. The growing medium was pre-washed potting sand (Fargro Ltd, Littlehampton, UK), with nutrients supplied in the form of the controlled release fertiliser Osmocote Mini Plus (Scotts Professional Limited, Nottingham, UK). Under the experimental conditions, Osmocote releases nutrients at a constant rate for nearly four months (Huett 1997; Day 2001). Nutrient-rich patches would therefore have supplied nutrients throughout the 60-day experimental period, rather than becoming depleted. Fertiliser was applied in either spatially heterogeneous or homogeneous arrangements. The total amount of fertiliser provided was dependent upon treatment. The same numbers of plants were sown, in identical spatial arrangements, in all treatments.

In heterogeneous treatments a chequerboard pattern of nutrient-rich and nutrient-poor patches was created by using a three-dimensional metal grid to divide each of 45 boxes, each 25 × 25 × 8 cm in size, into eight rows of eight 3.1 × 3.1 cm cells. Each cell was allocated a unique number to define its location (see Fig. 1). The grid was placed in the planting box and each cell was filled with either a fertiliser/sand mix or sand only. The grid was removed before planting to allow nutrients to diffuse between cells, and roots to grow freely between cells, as they would under patchy field conditions (e.g. Wijesinghe & Hutchings 1997, 1999). Diffusion of nutrients (especially mobile nutrients such as nitrate) through the substrate could have altered the original distribution of nutrients between cells, and enriched some nutrient-poor cells, especially those adjacent to nutrient-rich cells. In contrast, greater water and nutrient uptake by the larger mass of roots in nutrient-rich cells would be expected to slow or even reverse the movement of nutrients out of nutrient-rich cells. Our experiment was designed to permit such processes to take place, rather than prevent them. The appearance of algal growth only on the surface of nutrient-rich cells in the heterogeneous treatments, and the greater biomass recovered from those cells (see Results) indicate that, because of the continuing controlled release of nutrients throughout the experiment, differences persisted between cells designated nutrient-rich and nutrient-poor in the heterogeneous treatments in the face of such processes.

Figure 1.

Diagrams of heterogeneous (a and b) and homogeneous (c) treatments. In heterogeneous treatments, dark grey squares indicate nutrient-rich cells and white squares indicate cells with no nutrients added. In homogeneous treatments mid-grey squares indicate cells with an intermediate nutrient concentration. Numbers indicate the location of cells of harvested plants. All dimensions are in cm.

There were two patch size treatments. In the large-patch treatment, sets of 16 adjacent cells in a 4 × 4 arrangement were filled with substrate of the same quality (patch size 12.5 × 12.5 cm). In the small-scale treatment, four adjacent cells in a 2 × 2 arrangement were filled with substrate of the same quality (patch size 6.25 × 6.25 cm, Fig. 1). Each heterogeneous treatment contained equal numbers of nutrient-rich and nutrient-poor cells occupying the same area and proportion (50%) of the experimental arena, and consisting of equal substrate volumes in total (Fig. 1). In the homogeneous treatment all cells were filled with the same fertiliser/sand mixture. The term ‘experimental arena’ is used to denote the filled box.

The amount of fertiliser in the nutrient-rich patches in the three heterogeneous treatments was either 0.25, 1 or 4 times the manufacturer's recommended application rate of 1500 mg Osmocote L−1 of sand. At an application rate of 1500 mg L−1 of sand, Osmocote releases compounds of N, P and K at rates of 3.05, 1.52 and 2.10 mg L−1 day−1, respectively (see Huett (1997) and Day (2001) for full details of nutrient release rates). Nutrient-poor cells received no added nutrients. In the homogeneous treatments the amount of fertiliser in every cell was either 0.125, 0.5 or twice the manufacturer's recommended application rate. Each of the nine treatments (homogeneous treatment and heterogeneous treatments at two scales, each at three nutrient levels) was replicated five times. The number of replicates was determined following a power analysis undertaken on a random sample of data from the replicates in the experiment described in Day et al. (2003a).

Seeds of C. hirsuta (Chambers Seeds Ltd, Wellingborough, UK) were germinated on vermiculite (William Sinclair Horticulture Ltd, Kings Lynn, UK). After 14 days, seedlings were selected for uniformity of root and shoot size. Sixty-four C. hirsuta seedlings were transplanted into each experimental arena, with one seedling placed centrally within each cell. Throughout the experiment water was provided as required. The experiment was laid out as a fully randomised block on a wire mesh bench.

Harvesting was carried out after 60 days, when flowering had begun in all populations. Each plant was cut at the point of emergence from the substrate, placed in a labelled bag, and the number of its cell recorded. The substrate from within each cell was then removed individually using a cutting tool, and the substrate and roots were separated by washing. The roots from each cell were placed in labelled bags and the cell number was recorded. All plant parts were oven-dried at 80 °C to constant weight, and weighed. Plants growing in the outer cells of the experimental arenas were discarded, and further analysis was carried out on the above-ground parts of the remaining 36 plants and the roots harvested from each of the remaining 36 cells (Fig. 1).

Data analysis

All data were analysed using the statistical package SPSS version 8.0 for Windows (Generalized linear model procedure SPSS Inc., 1989–97). Generally, data were found to be normally distributed and homoscedastic.

hypothesis 1: the effects of nutrient availability and pattern of nutrient supply on population yield

For each treatment, the above- and below-ground biomass measurements from all 36 harvested cells were summed to determine population shoot mass, root mass and total mass. The effects of treatment on yield were analysed by two-way multivariate analyses of variance (manova) with nutrient level and pattern of nutrient supply as independent variables and population shoot and root mass as dependent variables. This was followed by the corresponding univariate analyses of variance (anova) to examine the effects of treatment on shoot and root mass. The effect of treatment on total population yield was investigated using two-way anova with nutrient level and pattern of nutrient supply as independent variables, and total biomass as the dependent variable. In all cases, significant differences between means were determined by Bonferroni multiple-means comparison tests at P < 0.05.

hypothesis 2: the effects of nutrient availability and pattern of nutrient supply on variation in individual shoot mass and root biomass per unit of substrate

Within-population variation in shoot mass of living plants and root biomass per cell was analysed by calculating their coefficient of variation (CV). Root biomass per cell is not equal to root biomass per plant, because the roots in each cell may have been produced by more than one plant. This value measures uniformity of substrate occupation by the roots of plants in the population. CVs from different treatments were compared using two-way manova with nutrient availability and pattern of nutrient supply as independent variables, and population shoot CV and CV of root biomass per cell as dependent variables. This was followed by the corresponding univariate anovas. When F-values were significant, results were subjected to Bonferroni multiple-means comparisons to determine which means differed significantly.

hypothesis 3: the effects of nutrient availability and pattern of nutrient supply on the spatial pattern of biomass distribution

To determine whether pattern of nutrient supply affected the spatial pattern of biomass distribution, the proportion of population biomass harvested from nutrient-rich patches was compared to an expected value of 50% (nutrient-rich patches occupied 50% of the experimental arena, both by area and by volume), using the one-sample t-test procedure. Additional one-sample t-tests were conducted on the homogeneous treatments. For these, random samples of 50% of the 36 harvested cells were taken. The proportion of the population biomass recovered from these cells was calculated, and compared with the expected value of 50%.

hypothesis 4: the effects of nutrient availability and pattern of nutrient supply on mortality

The effects of treatment on mortality were investigated using two-way anova with nutrient availability and pattern of nutrient supply as independent variables, and number of plant deaths as the dependent variable. Significant differences between means were determined using a Bonferroni multiple-means comparison test at P < 0.05.

Results

population yield

manova revealed a strong effect of nutrient availability and its distribution on population shoot and root biomass (Table 1). The corresponding anovas showed that population root, shoot and total biomass increased as nutrient availability increased (Table 1), as did total biomass (Table 2). Although the pattern of distribution of nutrients alone did not significantly affect these measures of yield, they were strongly affected by the interaction between availability of nutrients and the pattern of nutrient supply (Table 2). At the high nutrient level, population root biomass was significantly greater when nutrients were distributed homogeneously rather than heterogeneously. At other nutrient levels, yields did not differ significantly between treatments with different patterns of nutrient supply (Tables 1 and 2, Fig. 2).

Table 1.  The effect of nutrient availability and pattern of nutrient supply on population above- and below-ground biomass. The results of (a) two-way multivariate anovas with population shoot and root biomass as the dependent variables and nutrient availability (N) and pattern of nutrient supply (P) as independent variables; and (b) the corresponding univariate analyses of variance. d.f. = degrees of freedom, SS = sum of squares and MS = mean square. See Fig. 2 for means and standard errors
(a) SourceWilks’ lambdad.f.F-valueP
Nutrient availability (N)0.0594,7054.324< 0.001
Pattern of nutrient supply (P)0.7394,70 2.861    0.030
N × P0.4828,70 3.850    0.001
(b) SourceVariabled.f.SSMSF-valueP
NPopulation shoot biomass 2261.610130.805108.528< 0.001
Population root biomass 2  9.550  4.775 80.316< 0.001
PPopulation shoot biomass 2  2.150  1.075  0.892    0.419
Population root biomass 2  0.292  0.146  2.459    0.100
N × PPopulation shoot biomass 4 17.858  4.464  3.705    0.013
Population root biomass 4  2.042  0.510  8.585< 0.001
ErrorPopulation shoot biomass36 43.390  1.205  
Population root biomass36  2.140  0.059  
Table 2.  The effects of nutrient availability and pattern of nutrient supply on total population biomass. The table shows the results of two-way univariate anovas with total population biomass as the dependent variable and nutrient availability (N) and pattern of nutrient supply (P) as independent variables. d.f. = degrees of freedom, SS = sum of squares and MS = mean square. See Fig. 2 for means and standard errors
Sourced.f.SSMSF-valueP
N 2381.614190.80785.776< 0.001
P 2  3.552  1.776 0.798    0.458
N × P 4 29.393  7.348 3.304    0.021
Error36 80.082  2.224  
Figure 2.

Total population yield (complete bars) and population shoot (open bars) and root (filled bars) biomass in homogeneous and heterogeneous treatments at low (L), medium (M) and high (H) nutrient supply. Error bars show 1 SE of total (upward bars) and component means (downward bars). Letters indicate significant differences in yield measurements (different fonts refer to different yield components) between treatments at the P < 0.05 level, following Bonferroni multiple-means comparison tests. See Tables 1 & 2 for associated analyses.

population size variation

Multivariate anovas revealed that nutrient availability and its pattern of distribution had significant effects on CV of shoot biomass and CV of root biomass per cell within populations (Table 3). Nutrient supply had a significant effect on variation in shoot biomass but did not affect variation in root biomass per cell, whereas the pattern of distribution of nutrients had a significant effect on variation in root biomass per cell but did not affect variation in shoot biomass (Table 3, Fig. 3a,b). CV of shoot mass tended to be higher when nutrient supply was higher, but showed little difference between treatments with different patterns of nutrient supply at the same nutrient level. CVs of root biomass per cell tended to be higher in populations in heterogeneous than homogeneous conditions (Fig. 3a,b).

Table 3.  Analysis of the effects of nutrient availability and pattern of nutrient supply on within-population coefficient of variation (CV) in plant biomass. The table shows the results of (a) two-way multivariate analyses of variance, with within-population CV in shoot biomass and root biomass per cell as the dependent variables and nutrient availability (N) and pattern of nutrient supply (P) as independent variables; and (b) the corresponding univariate analyses of variance. d.f. = degrees of freedom, SS = sum of squares and MS = mean square. See Fig. 3 for means and standard errors
(a) SourceWilks’ lambdad.f.F-valueP
N0.4674,70 8.100< 0.001
P0.3674,7011.371< 0.001
N × P0.7668,70 1.250    0.284
(b) SourceVariabled.f.SSMSF-valueP
NWithin-population variation in shoot biomass 210324.9355162.467 7.396    0.002
Within-population variation in root biomass per cell 2  688.519 344.260 1.325    0.278
PWithin-population variation in shoot biomass 2  328.629 164.315 0.235    0.791
Within-population variation in root biomass per cell 2 9002.6814501.34117.324< 0.001
N × PWithin-population variation in shoot biomass 4 3935.588 983.897 1.410    0.250
Within-population variation in root biomass per cell 4  828.026 207.007 0.797    0.535
ErrorWithin-population variation in shoot biomass3625127.242 697.979  
Within-population variation in root biomass per cell36 9353.794 259.828  
Figure 3.

The effects of treatment on mean (+ 1 SE) within-population coefficient of variation in shoot biomass (a) and root biomass per cell (b) at low (L), medium (M) and high (H) levels of nutrient supply. Letters above the error bars indicate significant differences between treatments at the P < 0.05 level, following Bonferroni multiple-means comparison tests. See Table 3 for associated analysis.

pattern of biomass distribution

Figure 4 shows the mean biomass (shoot biomass plus cell root biomass) recovered from each harvested cell in treatments with different levels and patterns of nutrient supply. Yields per cell increased with nutrient supply, and, in heterogeneous treatments, were higher in nutrient-rich cells than in nutrient-poor cells. Visual inspection of Fig. 4 suggests that, in the large-scale heterogeneous treatment, yields in nutrient-poor cells adjacent to nutrient-rich cells were greater than in other nutrient-poor cells. This may have been a result of diffusion of nutrients from rich cells into poor cells, or roots of plants in poor cells growing into adjacent nutrient-rich patches. Whereas nutrient-rich patches made up half of the substrate area and volume in the heterogeneous treatments, significantly more than 50% of population biomass was located in these patches (Table 4, Fig. 5). In the heterogeneous treatments, mean values ranging from 68% (intermediate scale heterogeneity, low nutrient supply) to 80% (intermediate scale heterogeneity, high nutrient supply) of total population yield was recovered from the nutrient-rich patches. The proportion of biomass in nutrient-rich patches did not vary with patch scale.

Figure 4.

Mean biomass per cell (the sum of shoot and root biomass recovered from the cell) in the treatments. The vertical axis shows mean cell biomass (g). In heterogeneous treatments, dark squares indicate nutrient-rich cells and light squares indicate cells with no nutrients added. In homogeneous treatments mid-grey squares indicate cells with an intermediate nutrient concentration. The figures within the top panel at the left refer to specific cell locations, enabling orientation between this figure and Fig. 1.

Table 4.  Analysis of the effect of treatment on the proportion of biomass located in nutrient-rich patches in heterogeneous treatments and in 50% of cells selected at random in the homogeneous treatment. The table shows a summary of nine independent one-sample t-tests designed to test whether biomass distribution was spatially uniform for each treatment. d.f. = degrees of freedom. See Fig. 5 for means and standard errors for heterogeneous treatments and an indication of pattern of biomass distribution in the homogeneous treatments
Pattern of nutrient supplyNutrient availabilityd.f.t-valueP
Large scale heterogeneityLow4  14.940< 0.001
Medium4  15.008< 0.001
High4  11.287< 0.001
Intermediate scale heterogeneityLow4   8.467< 0.001
Medium4   9.968< 0.001
High4  10.842< 0.001
HomogeneousLow4−0.221    0.836
Medium4   0.377    0.725
High4−0.218    0.838
Figure 5.

Mean (+ 1 SE) proportion of total population biomass located in nutrient-rich cells in heterogeneous treatments, and in 50% of the cells, selected at random, in the homogeneous treatments at low (L), medium (M) and high (H) levels of nutrient supply. The dashed horizontal line indicates the predicted proportion of biomass in nutrient-rich cells if distribution is random. Letters above the error bars indicate significant differences between means at the P < 0.05 level following Bonferroni multiple-means comparison tests. See Table 4 for associated analysis.

Frequency histograms of shoot mass and root mass per cell, summed across all replicates for each combination of nutrient level and pattern of nutrient supply, are shown in Fig. 6. Dead plants are represented in the shoot weight histograms as having nominal, very low weights. At low nutrient levels, shoot size frequency distributions were unimodal and somewhat symmetrical for all patterns of nutrient supply, whereas at higher nutrient levels the distributions were positively skewed when nutrient supply was heterogeneous. The distribution of shoot sizes of surviving plants in the homogeneous treatment remained roughly symmetrical in the high nutrient treatment, but inclusion of dead plants with nominal low weights produced a frequency distribution with two distinct peaks. There was a clear disjunction between a large number of plants with extremely small shoots, many of which had died, and larger surviving plants with markedly higher, and widely ranging, shoot sizes (Fig. 6a). The populations in the heterogeneous treatments contained a smaller proportion of dead plants with small shoots, and the surviving plants showed a more continuous decline in frequency as shoot size increased.

Figure 6.

Frequency histograms of shoot weights (a) and root weights per cell (b). For ease of comparison between treatments, all graphs are drawn with 20 size classes on the x-axis, each 0.05 g in width in (a) and 0.005 g in width in (b). In the graphs for shoot weights, dead plants (black columns) are shown as having weights < 0.05 g. The y-axis represents number of observations in each weight class. Values have been pooled (n = 180) for the five replicate populations per treatment.

The frequency distributions of root biomass per cell revealed marked differences between treatments with homogeneous and heterogeneous nutrient supply (Fig. 6b). In homogeneous conditions at low nutrient supply the frequency distribution was unimodal, with all cells containing low root biomass, whereas in heterogeneous conditions the distribution was bimodal, with low root biomass in nutrient-poor cells and high root biomass in nutrient-rich cells. As nutrient supply increased, the frequency distribution in the homogeneous treatment remained unimodal, and no cells contained low root mass, particularly in the high nutrient treatment. In the heterogeneous treatments, the frequency distributions remained bimodal. Nutrient-rich cells again contained high root masses, but these values in general were no greater than the root mass in cells in the homogeneous treatment, although the nutrient-rich cells in the heterogeneous treatments contained more nutrients. Nutrient-poor cells in heterogeneous treatments contained little root biomass.

mortality

Mortality was significantly affected by level and pattern of nutrient supply. More plants died at higher levels of nutrient supply, and in homogeneous rather than heterogeneous treatments at all levels of nutrient supply (Table 5, Fig. 7). This difference was most marked at the low nutrient level, where populations in homogeneous conditions suffered at least eight times as many deaths as populations in heterogeneous conditions. In heterogeneous treatments there was a tendency for more plants in nutrient-rich patches to die (Fig. 7).

Table 5.  The effects of nutrient availability and pattern of nutrient supply on mortality in populations. The table shows the results of two-way univariate anovas with mortality as dependent variable and nutrient availability (N) and pattern of nutrient supply (P) as independent variables. d.f. = degrees of freedom, SS = sum of squares and MS = mean square. See Fig. 7 for means and standard errors
Sourced.f.SSMSF-valueP
N 2104.578 52.289 23.768< 0.001
P 2460.578230.289104.677< 0.001
N × P 4  5.556  1.389  0.631    0.643
Error36 79.200  2.200  
Figure 7.

Mean (+ 1 SE) number of plants dying in populations in heterogeneous and homogeneous treatments grown in low (L), medium (M) and high (H) levels of nutrient supply. In the heterogeneous treatments the dark columns show the mean number of deaths occurring in nutrient-rich cells and the open columns show the number of deaths occurring in nutrient-poor cells. Letters above the error bars indicate significant differences between means at the P < 0.05 level following Bonferroni multiple-means comparison tests. See Table 5 for associated analysis.

Discussion

The major results of this study were as follows. (i) As predicted by Hypothesis 1, populations with the same nutrient supply had similar yields at the time of harvest, regardless of pattern of nutrient supply. Populations grown under higher nutrient levels had higher yields. (ii) CV of shoot sizes was greater at higher levels of nutrient supply, but unaffected by pattern of nutrient supply. CV of root biomass per unit of substrate was significantly affected by pattern of nutrient supply but unaffected by availability of nutrients. The distributions of shoot sizes and of root mass per unit of substrate were strikingly different in populations growing at the same levels of nutrient supply but with different patterns of nutrient supply. Variation in shoot mass and root mass per cell was greater in heterogeneous conditions, especially at high nutrient supply. Thus, there was support for the predictions of Hypothesis 2. (iii) Although the biomass of populations growing under heterogeneous conditions was significantly concentrated on nutrient-rich patches, the proportion found in nutrient-rich patches did not vary with patch scale. Thus, there was support for Hypothesis 3. (iv) Contrary to the prediction of Hypothesis 4, however, mortality was greater in populations growing under homogeneous rather than heterogeneous conditions, at all nutrient levels, regardless of patch scale.

Day et al. (2003a) observed significant yield benefits after 31 days in populations of Cardamine hirsuta grown under heterogeneous conditions compared with populations in homogeneous conditions with the same overall level of nutrient supply. Many studies have shown that plants select more favourable patches of heterogeneous substrates in which to place their roots, and that significant root proliferation occurs in nutrient-rich patches, sometimes at the expense of root growth in nutrient-poor patches (e.g. Drew 1975; Jackson & Caldwell 1989; Gersani & Sachs 1992; Robinson 1994; Wijesinghe & Hutchings 1997, 1999; Fransen et al. 1998; Einsmann et al. 1999; Robinson et al. 1999). Plants that display such selectivity in root location, and root proliferation in more favourable patches of substrate, enhance their acquisition of nutrients, particularly in the presence of competitors (Robinson et al. 1999). Plants competing for the same pool of nutrients have also been shown to increase the proportion of their biomass allocated to roots at the expense of allocation to other functions, such as reproduction (Gersani et al. 2001). Acquisition of nutrients can be achieved more efficiently when the same amount of nutrients is concentrated into a smaller volume of the available substrate than when it is uniformly distributed (Kovar & Barber 1989; Jackson & Caldwell 1996). All of these facts suggest that essential resources could be acquired faster under heterogeneous conditions, resulting in greater growth, as in the studies cited above, and in Day et al. (2003a), at least until the supply of resources becomes exhausted.

Not all studies of population growth, however, have revealed increased growth under heterogeneous conditions (Casper & Cahill 1996). Possible reasons are that the scale of heterogeneity used may have been inappropriate for the species investigated to respond to (Wijesinghe et al. 2001; Hutchings et al. 2003), or that growth had become resource-limited. It would be predicted from the marginal value theorem (Charnov 1976) that when growth had become resource-limited, differences in resource availability, both between and within treatments, would have been eliminated. Moreover, regardless of any ability of populations to develop at different initial rates under homogeneous and heterogeneous conditions, yields would converge. This is what happened in the current experiment, in which C. hirsuta was grown for twice as long as in Day et al. (2003a) (see also Fransen & de Kroon 2001). Although populations achieved similar yields, however, they retained striking differences in structure and developed different patterns of mortality.

Coefficients of variation in shoot weights generally revealed few differences between populations in heterogeneous and homogeneous conditions at the same level of nutrient supply (Fig. 3a), despite values having been greater in heterogeneous conditions after 31 days of growth (Day et al. 2003a; see also Fransen et al. 2001). In contrast, shoot weight frequency distributions were strongly affected by both level and pattern of nutrient supply. The differences were greater still when dead plants were included in the histograms (Fig. 6a). Under homogeneous conditions the surviving plants had approximately symmetrical shoot weight distributions, with few small living plants, but a large number of dead plants. Under heterogeneous conditions the shoot weight distributions of living plants were skewed towards the smallest size classes, but fewer plants had died. Coefficients of variation for root biomass within cells were low in the homogeneous treatments compared with the heterogeneous treatments (Fig. 3b), and the frequency histograms of the data showed strong single peaks, indicating that most cells contained similar quantities of root (Fig. 6b). Few cells contained large or small root masses. The histograms of root biomass within cells were strongly bimodal in all heterogeneous treatments, reflecting the high or low nutrient concentration in individual cells. The nutrient-rich cells in these treatments had a much higher nutrient content than any cells in the homogeneous treatment at the same level of nutrient supply. However, at the higher nutrient levels they did not contain more root than cells in the homogeneous treatment. This may indicate either that there is a physical limit to the amount of root that can be packed into a volume of substrate (e.g. McConnaughay & Bazzaz 1992) and that this value cannot be exceeded even if nutrient concentration is raised, or that efficiency of nutrient acquisition is maximized when this quantity of root has been packed into a given volume of substrate (e.g. Robinson 1996).

The high root mass in nearly all cells in the homogeneous treatments would be expected to result in intense exploitation of soil-based resources throughout the whole substrate. In contrast, a large number of cells in the heterogeneous treatments, mostly with nutrient-poor substrate, contained low root weights, suggesting that these cells might be refuges from intense below-ground competition because they were less heavily exploited. Although plants located in these cells were not large, and may not have been able to project roots into nutrient-rich patches, even though such patches were very close (Hutchings et al. 2003), they had a lower probability of dying than plants in nutrient-rich cells (Fig. 7). Competition for soil-based resources, and perhaps for space, is likely to have been more intense in nutrient-rich cells in the heterogeneous treatments. However, it may have been no more intense than in the homogeneous treatment, as root masses per cell were generally similar. Moreover, competition for light may have been less severe for plants located in nutrient-rich cells that were adjacent to smaller plants in nutrient-poor cells. In summary, homogeneous conditions appear to be characterized by uniformly strong exploitation of resources below, and possibly also above ground. In comparison, in heterogeneous conditions, intense exploitation of soil-based resources, and possibly also light, appears to be confined to a smaller proportion of the habitat, causing strong inter-plant competition to be more localized in occurrence. Ultimately, there was more mortality under homogeneous conditions as a consequence of strong competitive interactions throughout the population. Casper & Cahill (1996) have also reported significantly higher mortality when populations of Abutilon theophrasti were grown under homogeneous conditions than under heterogeneous conditions providing the same overall level of nutrients, suggesting that this may be a general result.

There is an extensive literature on the structure of plant populations (e.g. Weiner & Thomas 1986; review in Hutchings 1997). Most experimental studies on which our understanding is based have been carried out under homogeneous conditions, either in the glasshouse or under highly artificial field conditions. However, spatial and temporal heterogeneity in the availability of essential resources is ubiquitous in natural environments (Jackson & Caldwell 1989, 1993; Caldwell et al. 1991; Levin 1992; Caldwell & Pearcy 1994; Stark 1994; Richard et al. 2000), suggesting that analysis of the development of population structure under heterogeneous conditions will deepen our understanding of the behaviour of populations under natural conditions. The results of this study reveal that subjecting plant populations to heterogeneous environmental conditions adds a new level of complexity to the processes that shape their structure. Importantly, this study, together with Day et al. (2003a), demonstrates that pattern of resource supply affects many facets of the performance of plant populations and of their component plants, and that the responses change in time in predictable ways that appear to depend on local patterns of resource utilization.

Acknowledgements

We gratefully acknowledge practical help from David Aplin during this experiment, and valuable comments on an earlier draft of the manuscript from David Gibson, Sue Hartley, David Robinson and four anonymous referees. This study was supported by a postgraduate studentship awarded to KJD by the University of Sussex.

Ancillary