The responses of seven co-occurring woodland herbaceous perennials to localized nutrient-rich patches

Authors


Rebecca Farley, University of Liverpool, School of Biological Sciences, Nicholson Building, Liverpool L69 3BX, UK (fax 0151 7945094; e-mail RAF4@liverpool.ac.uk).

Summary

1 Nutrient-rich patches can occur in soils at a wide range of spatial and temporal scales. Differences in the root proliferation response between species may be due to differing abilities of root systems to locate and recognize patches of differing size and nutrient concentration.

2 We investigated the root proliferation responses of seven co-occurring herbaceous perennial British woodland species (Ajuga reptans, Glechoma hederacea, Oxalis acetosella, Silene dioica, Stachys sylvatica, Veronica montana and Viola riviniana) and the effect of mycorrhizal colonization on any response.

3 Plants were planted in nutrient-poor sand in the centre of rectangular pots, with a nutrient-rich patch to one side and a control (sand) patch on the other. Size and nutrient concentration of the patches were varied between treatments.

4 Species differed in the size of their root systems and in their ability to respond to localized nutrient-rich patches. Oxalis acetosella and Viola riviniana, which produced the smallest root systems, showed similar root growth in nutrient-rich and control patches. All other species responded to the presence of a nutrient-rich patch by various combinations of root proliferation, changes in root branching pattern and by an increase in specific root length.

5 In some species the response was affected by patch attributes: Silene dioica and Veronica montana were sensitive to the nutrient concentration of the patch, and Glechoma hederacea did not respond to the smallest sized patch.

6 Mycorrhizal colonization had little effect on root proliferation. Only one species (Oxalis acetosella) could be shown to benefit from colonization by increased phosphate uptake.

7 The soil is a heterogeneous environment in terms of nutrient availability; differences between species in the ability to exploit this heterogeneity may affect their distribution, and could be a mechanism that reduces interspecific root competition.

Introduction

Soil is a heterogeneous environment in which water and nutrient concentrations vary both temporally and spatially. Nutrient heterogeneity occurs at a wide range of scales, including those relevant to the root systems of individual plants (Lechowicz & Bell 1991; Jackson & Caldwell 1993; Robertson et al. 1993; Farley & Fitter 1999). Such spatial variation creates a mosaic of patches differing in size and nutrient availability. Therefore, a plant root is likely to experience a range of nutrient concentrations as it grows through the soil. Roots of many species are known to proliferate lateral branches in more fertile areas of soil (Sprague 1933; Drew & Saker 1975) and, although the degree of this response differs (Robinson 1994), it is not known what determines such interspecific differences.

Campbell et al. (1991) found that species with large root systems had a lower proportion of their roots in nutrient-rich patches than did small species. Thus, although these patches still contain more roots of large species, these species appear to be less effective at localized, precise foraging. They proposed a trade off between scale and precision such that plant species that foraged on a smaller scale showed greater precision in their foraging.

Species may differ in the types of patch to which they can respond by root proliferation. Given that the roots of co-occurring plants can intermingle (Pecháckováet al. 1999), if species with coarse root systems forage widely but insensitively they may leave small nutrient-rich areas that precise foragers can exploit, and such differentiation might enable plants to avoid competition for nutrients. However, the species used by Campbell et al. (1991) were unlikely to co-occur, and the hypothesis has yet to be tested on species exploiting a common pattern of soil heterogeneity under natural conditions.

The potential advantage of being able to exploit nutrient-rich patches by localized proliferation may depend on the nature of the root system. Coarse roots of low specific root length (SRL; length of root per unit weight) are more expensive for a plant to produce than fine ones, as they have a higher carbon cost per unit root length, and coarse-rooted plants may therefore gain little benefit from such a response. Root architecture, i.e. how the roots branch and how they are arranged in the soil, can affect the efficiency of nutrient acquisition (see models in Fitter et al. 1991). For instance, although herringbone branching is more expensive to produce per unit length, it gives increased efficiency in nutrient-poor soils and may not be favoured in nutrient-rich patches.

Proliferation responses may also be affected by associated species such as mycorrhiza. Although pot experiments show that arbuscular mycorrhiza can enhance phosphate uptake and improve plant growth, the only temperate and alpine species shown to exhibit enhanced phosphate uptake when mycorrhizal under field conditions have very restricted root systems, such as those of the bluebell, Hyacinthoides nonscripta (Merryweather & Fitter 1995) and or the alpine buttercup, Ranunculus adoneus (Mullen & Schmidt 1993). If, however, plants depend on mycorrhiza for nutrient uptake, colonization might reduce root proliferation in nutrient-rich patches.

Seven herbaceous perennials that co-occur naturally at a woodland site in North Yorkshire, UK, were studied for their response to nutrient-rich patches. This woodland has a nutrient-poor soil (Merryweather & Fitter 1995) that shows a high degree of spatial and temporal heterogeneity in soil nutrients, with short-lived nutrient-rich patches occurring at a scale relevant to plant root systems (Farley & Fitter 1999). The native plants may therefore be expected to show an ability to exploit nutrient-rich patches, but any morphological responses would need to be expressed rapidly. The objectives of the experiment were to determine:

1 whether this group of co-occurring species could respond morphologically to nutrient-rich patches of differing size and nutrient concentration during a short-term experiment, and whether responsiveness was a function of SRL; we predicted that species with low SRL would be less responsive to nutrient-rich patches than species with high SRL;

2 whether mycorrhizal colonization altered any response to nutrient patches; we predicted that species that are most dependent on mycorrhiza would be least likely to show a response to nutrient-rich patches;

3 whether any response to nutrient-rich patches or mycorrhizal colonization was accompanied by changes in SRL or underground architecture; we predicted that responsive species should produce fine roots (higher SRL) and more dichotomously branched root systems in nutrient-rich patches.

Methods

Experiment 1

Seven herbaceous perennial dicotyledonous species (Table 1) that co-occur at Pretty Wood, Castle Howard, North Yorkshire, UK (National Grid reference SE 732 687), were used. Plants were collected from the site in October 1992 and propagated in a glasshouse, except for Silene dioica which was grown from seed (Emorsgate Seeds, Norfolk, UK).

Table 1.  Species used in the experiment, their mycorrhizal status and typical height
SpeciesFamilyNormal mycorrhizal status*Plant height (cm)†
  1. *Harley & Harley (1987). †Fitter & Peat (1994)

Ajuga reptans L.LamiaceaeMycorrhizal5–30
Glechoma hederacea L.LamiaceaeMycorrhizal5–30
Oxalis acetosella L.OxalidaceaeMycorrhizal5–10
Silene dioica (L). Clairv.CaryophyllaceaeNon-mycorrhizal30–90
Stachys sylvatica L.LamiaceaeMycorrhizal30–100
Veronica montana L.ScrophulariaceaeMycorrhizal5–15
Viola riviniana Reichb.ViolaceaeMycorrhizal5–15

Small plantlets with roots and at least four leaves were transplanted to the centre of rectangular pots, 20 cm long by 5 cm wide and 4 cm deep, which were subdivided using coarse (2 mm) netting (Fig. 1) to help maintain distinct patches and ease harvesting without impeding root growth. Each pot had a nutrient-enriched patch on one side of the central plant, and a un-enriched area of the same size on the other side to serve as a control patch (Fig. 1). The patches began 2 cm from the plant, so that there was an equal chance of roots finding and exploring either patch if they grew concentrically from the plant. At the start of the experiment the roots of all species were restricted to the central area, and any found outside this area at harvest represented new growth.

Figure 1.

Diagram showing the division of the pots into patch and control areas for (a) small-, (b) medium- and (c) large-sized patches.

Three patch sizes were created by varying the length of the patch (Fig. 1). The small patch extended for 2 cm (volume 40 cm3), with the remaining 6 cm of the pot being filled with sand. The medium patch extended for 3.5 cm (volume 70 cm3), with the remaining 4.5 cm of the pot being filled with sand. The large patch extended for 8 cm (volume 160 cm3), to the end of the pot. Coarse washed sand was used in all areas except the nutrient-rich patch, which was filled with either 100% soil or a 50% soil/50% sand mix (subsequently measured nutrient levels are given in Table 2). Soil was obtained from a garden at the University of York. Three replicates for each combination of species, patch size and patch concentration were established in December 1993 and a further three replicates of each treatment were started in March 1994. Plants were harvested after 6 weeks of growth in a heated glasshouse (temperature range 13–20 °C, with supplementary sodium lighting of 250 µmol m–2 s–1 at plant height for a 16-h day).

Table 2.  Ammonium, nitrate and phosphate concentrations (µmol g–1 dwt) of the sand and nutrient-enriched soils. Concentrations measured in Pretty Wood, recalculated from Farley &; Fitter (1999), with their max:min ratio in parentheses, are included for comparison
 NH4+µmol g–1 dwtNO3µmol g–1 dwtP µmol g–1 dwtMean ratio patch : sand for concentration of all nutrients
Sand0.070.050.011 : 1
50% soil patch0.380.190.179 : 1
100% soil patch0.690.340.4719 : 1
Pretty Wood0.15–2.22
(14 : 1)
0.15–1.91
(12 : 1)
0.05–0.42
(8 : 1)
 

After the shoots had been harvested, the soil was divided along the patch boundaries, cutting the roots, and the roots were extracted from each section. Root length was calculated using the program fibre (version 4.4) on an Applied Imaging Magiscan Image Analyser (applied Imaging, Gateshead, UK). The two largest root fragments were selected from both the enriched and control patches of the large patch treatment and spread on glass plates so that their architecture could be measured using the program trackroot (Fitter 1993) on the same Image Analyser. As root length was dependent on the volume of soil from which roots were extracted (patch size), results were expressed as root length densities (km m–3). Response to the nutrient-rich patch was expressed as the difference in root length density between the control and nutrient-enriched patch in each pot. Precision of the root proliferation response was measured as the difference between root length density in the enriched and control patches divided by the root length density in the control patch; a value of zero indicate no precision in placement of roots.

Experiment 2

The plants used for experiment 1 had been propagated in compost so that they were not mycorrhizal. Although the nutrient-enriched patches contained unsterilized soil, they contained little or no mycorrhizal inoculum and the roots did not become mycorrhizal during the course of the experiment. Under field conditions all but one (Silene dioica) of the selected species are to some extent mycorrhizal (Table 1).

To determine whether mycorrhizal colonization affected patch exploration by roots, six replicate pots were set up for each species in December 1994. All had a medium-sized (3.5 cm long) nutrient-rich patch on either side of the plant (both patches filled with soil). One of the nutrient-rich patches was supplemented with an inoculum of Glomus sp. BEG 6 [as 0.93 g soil containing colonized fragments of Plantago lanceolata roots per 48 g of soil (fresh weight)]. Plants were harvested at 10 weeks, as mycorrhizal colonization was poorly established at 6 weeks. Root systems were extracted from each patch, each sample was divided into two equal portions, and both were weighed. One portion of root was used to estimate length of the whole root sample and the two largest fragments were used for architectural measurements, as above. These portions were then dried and weighed.

The second portion was used to determine the arbuscular mycorrhizal colonization of the roots. Roots were stained for mycorrhiza in acid fuchsin (modified from Kormanik & McGraw 1982 to omit phenol) and colonization was determined using epifluorescence microscopy (Merryweather & Fitter 1991) at 250 × magnification using 100 intersections per slide (McGonigle et al. 1990).

Phosphate concentration of the shoots was determined by acid digestion of the dried tissue and analysed at 882 nm using ascorbic acid-molybdenum blue (Allen 1974). Plants from the mycorrhizal experiment were compared with the plants from the main experiment with large 100% concentration patches, as these pots contained approximately the same volume of soil in one nutrient-enriched patch as those in the mycorrhizal experiment did in two. Therefore the plants had similar quantities of nutrients available for growth.

Statistical analysis

Root architecture was measured using a topological index (Fitter 1993), by regression of log (pathlength) on log (magnitude), using Minitab, release 9. Pathlength is the total number of links in all unique paths from an exterior link to the origin of the root system, and magnitude is the number of exterior links (Fitter 1993). Differences in root system architecture due to nutrient enrichment were tested using Student's t-test, to seek differences between the regression coefficients (Steel & Torrie 1980). High slope coefficients occur with herringbone architectures, with a theoretical maximum slope of 1.96. Lower values represent more random or even dichotomous branching patterns.

All other data were subjected to either analysis of covariance or analysis of variance on Minitab, release 9. As the experiment had been run in batches, each batch of three replicates was treated as a block, but blocks never had a significant effect. Root fraction (root weight divided by plant weight), SRL and response to nutrient-rich patches were analysed by three-way analysis of covariance for species, patch size and patch concentration, with total plant root length as the covariate. Mycorrhizal colonization and shoot phosphate concentration were subjected to analysis of variance for species and mycorrhizal status, as the covariate (root length) had no significant effect on the results. Species were considered individually by one-way analysis, where appropriate, to examine the effects of the size or concentration of the nutrient-rich patch (experiment 1) and for the effect of mycorrhizal colonization (experiment 2).

Results

Experiment 1

Above-ground biomass varied sevenfold between species (Table 3). Ajuga reptans and Glechoma hederacea produced the most above-ground biomass, while Oxalis acetosella produced significantly less than all the other species. Species also differed in the total length of their root systems (Fig. 2) and were allocated to one of two groups according to the size of root system produced. Glechoma hederacea, Ajuga reptans, Veronica montana, Stachys sylvatica and Silene dioica produced large root systems (total root length 7–16 m plant–1) that extended throughout the pots (Fig. 3). Oxalis acetosella and Viola riviniana produced significantly smaller root systems (total root length 1–3 m plant–1), with roots predominately in the central area of the pot; only a small proportion of their roots entered the patches and few roots, if any, extended beyond the patches (Fig. 3). SRL was lower for Ajuga reptans and the two species with small root systems than the other species (Table 3).

Table 3.  Mean species characteristics. Above-ground dry weight (mg) and specific root length (SRL; m g–1 dwt) are expressed as means per pot across all treatments (± SE) and both show significant variation between species. Precision (± SE) in root foraging [(difference in root length density between the enriched and control)/control] did not differ between species. F- and P-values and least significant difference (LSD) are given where appropriate. Subscripts within a column show values that differ significantly
SpeciesMean above-ground dry weight (mg)Mean SRL (m g–1)Precision in foragingn
Ajuga reptans225 (± 20)c46.1 (± 4.0) a4.8 (± 1.8)24
Glechoma hederacea258 (± 23) c173.9 (± 8.7) c1.7 (± 0.4)24
Oxalis acetosella36 (± 4) a52.0 (± 4.0) aNA22
Silene dioica134 (± 16) b115.3 (± 8.2) b3.0 (± 1.6)24
Stachys sylvatica174 (± 29) bc87.7 (± 9.3) b2.0 (± 0. 8)21
Veronica montana152 (± 14) b148.7 (± 3.4) c1.1 (± 0.3)24
Viola riviniana121 (± 10) b35.7 (± 8.6) aNA20
F6,11717.8552.321.6 
P < 0.001 < 0.0010.18 
LSD68.332.0  
Figure 2.

Root system length (mean plant root length, m; open bars, right axis) and ability of species to respond to nutrient enrichment by root proliferation [expressed as the difference in root length density (km m–3) between enriched and control patches; closed bars, left axis]. Values are expressed as means for all treatments (± SE, n = 24). There was significant variation in total root length between species (F6,117 = 31.09, P < 0.001) and in their ability to respond to nutrient enrichment (F6,115 = 3.68, P = 0.002); significant responses (P < 0.05) to nutrient enrichment (i.e. significantly greater root density in nutrient-rich patches than in control patches) are indicated by *.

Figure 3.

Mean proportion of the root system in each part of the pot (± SE), expressed as a percentage of total root length, for pots with (a) small, (b) medium and (c) large patches. Post-control and post-patch refer to the areas distal to the control and nutrient-rich patches, respectively; in the large patch treatment there was no such area.

Species differed in their ability to respond to nutrient-rich patches (Fig. 2). A significant proliferation of roots in the nutrient-enriched patch was seen in four of the seven species (Ajuga reptans, Glechoma hederacea, Silene dioica and Veronica montana). An apparent response by the remaining large-rooted species, Stachys sylvatica, was obscured by high variation between replicates. Although roots of Oxalis acetosella and Viola riviniana grew out of the central patch (Fig. 3) there was no significant difference in root length density in the nutrient-rich and control patches. The mean response to nutrient enrichment was correlated with mean total root length (r5 = 0.73, P < 0.05). The precision of root foraging within the nutrient-rich patch did not differ significantly between the responsive species (Table 3) and could not be tested for the small rooted species where few roots extended into either the enriched or the control patch.

Patch nutrient concentration affected allocation of biomass of all species. Plants grown with the 50% patch allocated a significantly higher proportion of resources to root growth than those grown with 100% patches (Fig. 4). The proliferation response was also affected by the patch concentration (F1,115 = 8.22, P = 0.005), with a greater response generally seen in the stronger patch treatment. However, differences in responses to patch concentration were significant only for Silene dioica and Veronica montana (Fig. 5). Silene dioica responded markedly to 100% patches but not to the 50% patch, suggesting that it only responds when there is a large difference between the patch and background soil. Veronica montana responded to both patch concentrations, but showed a significantly greater response to the higher concentration. Ajuga reptans and Glechoma hederacea both responded equally to the two patch concentrations.

Figure 4.

Root fractions of whole plants for 50% and 100% nutrient enrichment treatments. Values are expressed as means for all patch sizes (± SE, n = 12). Mean root fraction was significantly greater for plants with a 50% soil nutrient patch than for plants with a 100% soil nutrient patch (concentration F1,115 = 1120.49, P < 0.001, species × concentration F1,115 = 7.42, P < 0.001).

Figure 5.

Proliferation responses to patch concentration, expressed as the difference in root density (km m–3) in enriched and control patches. Values are expressed as means for all patch sizes for a particular patch concentration (± SE, n = 12); a value of zero represents no response. Species that showed a significant response to nutrient enrichment are marked as A (response to 50% soil patch) or B (response to 100% concentration patch) and were not significantly different (at P < 0.05) if marked AB.

All the species with large root systems had roots that extended through the small- and medium-sized patches into the distal areas of the pots. Therefore the plants must have experienced differences in patch size in different treatments (Fig. 3). However, patch size had little effect on the root proliferation response (F2,115 = 2.95, P = 0.056). Only one species, Glechoma hederacea, was sensitive to patch size (Fig. 6), with root density in the large patch being lower than in the medium patch (8.9 and 5.4 km m–3, respectively) and no response to small patches. Therefore, even though Glechoma hederacea produced a large root system, it probably did not exploit fully the large patch during the 6 weeks of the experiment.

Figure 6.

Species’ response to patch size, expressed as the difference in root length density (km m–3) in enriched and control patches. Values are expressed as means for all patch concentrations (± SE, n = 8); a value of zero represents no response. Responses (A, small patch; B, medium patch; C, large patch) did not differ between patch sizes except for Glechoma hederacea, which showed a significant response to medium and large patches only, as shown by one-way anova (size F2,21 = 7.19, P = 0.004), with a greater response to medium than large patches (B > C).

There was a tendency for species with high SRL to be more responsive (Table 3; r156 = 0.34, P < 0.001, for all data). Plants produced thinner roots in the nutrient-enriched patch than in the control (F1,233 = 4.75, P = 0.03). When analysed by one-way analysis of variance, only Silene dioica and Ajuga reptans showed a significant amount of plasticity in SRL (Fig. 7).

Figure 7.

Specific root length (SRL, m g–1 dwt) for sections of the root system in enriched and control patches for all species, expressed as means over all treatments (± SE, n = 24). A significant difference (P < 0.05) in SRL between the two patches is indicated by *.

Most species that responded to soil nutrient heterogeneity by root proliferation achieved this partly by a change in root system architecture (Fig. 8). In the control patches, the higher topological index indicated a more herringbone-like structure compared with more randomly branched root systems in enriched patches. Stachys sylvatica, which did not respond significantly to nutrient enrichment by root proliferation (Fig. 2), showed the largest changes in root architecture. Changes in root architecture were more often significant in plants grown with a 50% concentration patch than with a 100% concentration patch. Glechoma hederacea was the only responsive species that showed no significant change in root architecture at either patch concentration.

Figure 8.

Root system architecture in enriched and control patches, with topological index calculated by the regression of log(pathlength) on log(magnitude). The dotted line represents the theoretical maximum slope of 1.96 for a herringbone structure. A decrease in topological index between control and enriched patches shows a change to more random branching in the enriched patch; a significant difference in branching pattern within treatments (P < 0.05) is indicated by *. Data are not available for Oxalis acetosella and Viola riviniana as root fragments were too small to analyse.

Experiment 2

When grown with a patch containing mycorrhizal inoculum, roots within the inoculated patch became colonized for all species (Table 4). Roots in the control patch showed no colonization except for Oxalis acetosella, where they were weakly colonized. Glechoma hederacea, Oxalis acetosella and Viola riviniana had a significantly greater colonization in the mycorrhizal than the control patch (P < 0.05). Analysis of tissue phosphate (Fig. 9) showed that inoculated plants of Oxalis acetosella had more than double the phosphate concentration of shoots of comparable non-mycorrhizal plants where shoot phosphate concentrations were extremely low. Mycorrhizal inoculation had no effect on root length, root system architecture or the ability to respond to patches by root proliferation. As the intensity of colonization was low, and only Oxalis acetosella benefited nutritionally from mycorrhizal treatment, the lack of such effects is perhaps not surprising.

Table 4.  Mean percentage mycorrhizal colonization for roots in control and mycorrhizal inoculated patches in experiment 2 (± SE). There was a significant difference in intensity of infection between control and inoculated patches (F6,35 = 3.41, P < 0.01). Species marked * had a significantly higher level of colonization in the mycorrhizal than control patch
SpeciesNon-mycorrhizal patchMycorrhizal patch
Ajuga reptans0 (± 0)10.0 (± 3.0)
Glechoma hederacea0 (± 0)14.0 (± 3.6)*
Oxalis acetosella2.0 (± 0.6)27.5 (± 1.3)*
Silene dioica0 (± 0)4.7 (± 2.7)
Stachys sylvatica0 (± 0)3.8 (± 1.4)
Veronica montana0 (± 0)6.5 (± 2.8)
Viola riviniana0 (± 0)8.5 (± 1.9)*
Figure 9.

Mean shoot phosphate concentration (mg g–1) of plants from the mycorrhizal experiment and of non-mycorrhizal plants from the 100%, large patches of experiment 1. There was a significant effect of inoculation on shoot phosphate concentration (species F6,42 = 9.55, P = 0.000, species × treatment F6,42 = 2.21, P = 0.06), but only Oxalis acetosella had significantly greater shoot phosphate concentration when mycorrhizal (*) than when non-mycorrhizal.

Discussion

Within this group of co-occurring species there were differences in both the magnitude and mechanism of response to localized nutrient-rich patches, and in the size and concentration of patches that species could exploit. Oxalis acetosella and Viola riviniana produced very small root systems and foraged for nutrients at a small scale. Although their roots entered the nutrient-rich patches, they showed no ability to exploit them by root proliferation. However, the proportion of their roots outside the central zone was lower than for the large rooted species, and given a longer time period it is therefore possible that these species might have responded to nutrient-rich patches. However, the nutrient-rich patches in the woodland where they occur are short-lived (Farley & Fitter 1999), and while seasonal differences in nutrient availability across the site may last for several months, small-scale nutrient-rich areas rarely last for more than 2–4 weeks. The time–course of the experiment therefore reflected this short-term nutrient availability.

Instead of a trade off between the scale and precision of resource foraging, as proposed by Campbell et al. (1991), precise foraging was found predominantly in species with large root systems that are capable of exploring large volumes of soil, as also seen by Einsmann et al. (1999). Moreover, although the length of the root system varied twofold (7–16 m) within this group of species, there was no significant difference in the precision with which they responded to nutrient-rich patches. As species with a large root system are likely to experience many different nutrient environments as they grow, an ability to exploit nutrient-rich patches by root proliferation would appear to be advantageous.

Root growth involves both construction and maintenance costs, and proliferation of roots increases these costs. For an optimal response to nutrient-rich patches, the cost of proliferation should be more than offset by the benefit gained from the patch. Some species showed an ability to adjust their root proliferation response according to the size (Glechoma hederacea; as also seen during clonal foraging Wijesinghe & Hutchings 1997) or concentration (Silene dioica and Veronica montana) of the nutrient-rich patch. Small patches and low concentration patches offer less return in terms of nutrients to the plant, and so the species that failed to respond (Glechoma hederacea and Silene dioica, respectively) or showed a reduced response (Veronica montana) may ensure a greater return for their investment of root growth into patches. If root proliferation in patches is modulated by the concentration of the patch (a response shown by Jackson & Caldwell 1989 and here by Veronica montana), the investment of root growth might lead to progressively greater benefit. The cost of root proliferation can also be reduced by producing thinner roots with lower construction costs. This plasticity may be especially important for coarse-rooted species such as Ajuga reptans, where the cost of proliferation would otherwise be high, compared with very fine-rooted species such as Silene dioica, where proliferating thinner roots further reduces the cost of exploitation of patches which is already low.

Plastic changes were seen in root architecture in response to nutrient-rich patches, and these fit the predictions of Fitter et al. (1991) for the types of root architecture that should be produced in uniform nutrient-rich or nutrient-poor soil. All the species produced an efficient but expensive (in terms of root construction costs) herringbone system in the low nutrient control patch, but less expensive more dichotomous branching in the nutrient-rich patch. The herringbone branching pattern is not suited to patch exploitation; branches are only formed off the main axis and the root therefore tends to grow through a patch before the patch has been fully exploited (Fitter 1994). The change in architecture thus both reduces proliferation cost and increases the efficiency of patch exploitation. Glechoma hederacea was the only responsive species that showed no plasticity in root architecture, while Stachys sylvatica was the only species to respond solely by a change in architecture. Exploitation of patches in Stachys sylvatica therefore depends on producing a more effective spatial pattern of roots in the patch, rather than any increase in their number or length.

Although mycorrhizal colonization did increase nutrient uptake in one species (Oxalis acetosella), there was no evidence that it altered the manner in which roots foraged for nutrients. Species with small root systems that are furthermore unable to exploit patchiness (like Oxalis acetosella and Viola riviniana) may face difficulties maintaining nutrient uptake, and mycorrhizal associations may therefore aid nutrient uptake as hyphae can explore a far greater volume of soil than the plant's own roots. Whether the mycorrhizal fungi themselves show a response to soil nutrient patches is still unclear. Although St John et al. (1983) showed that hyphae of mycorrhizal fungi proliferated in nutrient-enriched sites in laboratory experiments, Duke et al. (1994) found that mycorrhizal colonization of Agropyron desertorum and Artemisia tridentata did not increase in nutrient-rich patches, although this does not prove that the hyphae did not proliferate. The apparent lack of nutritional benefit from mycorrhizal colonization in Viola riviniana may be because only a low level of colonization was achieved during the experiment; at the field site it had a higher intensity of colonization (16%; R. A. Farley, unpublished data). Mycorrhizal colonization offered no nutritional benefit for the other species during the short time scale of this experiment. These species, which all responded to nutrient-rich patches, produced high root length densities and were presumably capable of supporting their nutrient requirements in the absence of mycorrhizal colonization. This may, however, offer alternative benefits, such as increased protection from root pathogens (Newsham et al. 1995).

Analysis of the differing responses of this group of co-occurring species suggests that those with copious roots are both more likely to encounter and better able to exploit nutrient-rich areas. The response may involve proliferation of roots (Glechoma hederacea) or a change in root architecture (Stachys sylvatica) or a combination of these (Ajuga reptans, Silene dioica, Veronica montana) and the size and concentration of exploitable patches varies between species. All the responsive species demonstrated a high level of plasticity in terms of the amount and type of roots that they produced, supporting De Kroon & Hutchings’ (1995) suggestion that morphological plasticity is important for the exploitation of localized resources.

These data demonstrate that there is wide variation in the degree and nature of plant root response to nutrient heterogeneity. They provide the first evidence that for some species that do not respond to patches at all by root proliferation, this may be due to the cost of proliferation in very coarse-rooted species or to their dependence on mycorrhizal associations. These data support those reviewed by Eissenstat (1992), which suggested that species that produce thin roots of high SRL may produce a more plastic root system than those of low SRL.

Environmental heterogeneity can affect the distribution of species; Breashears et al. (1997) demonstrated differential use of heterogeneously distributed soil water by co-existing shrubs, which was consistent with the distribution of the species within the habitat. Our results show that species that co-occur in a habitat, and which therefore are likely to encounter similar patterns of heterogeneity, differ both in the way in which they respond to patches and to the type of patch to which they respond. These differences may affect species’ distribution within the habitat and may be an element of the niche differentiation that permits co-existence.

Acknowledgements

Thanks are due to Castle Howard Estate for allowing us to collect samples from Pretty Wood, and to the Natural Environment Research Council for financial support. We thank M. Hutchings, B. Campbell and an anonymous referee for their suggestions on an earlier draft of this manuscript.

Received 29 October 1998revision accepted 24 March 1999

Ancillary