Morphological responses of plant roots to heterogeneity of soil resources


Author for correspondence: Marie Šmilauerová Tel: +420 38 7772325Fax: +420 38 5300366 E-mail:


• Root morphological response to experimentally induced soil heterogeneity is reported here on three grassland species (Luzula campestris, Poa angustifolia and Plantago lanceolata) under field conditions.

• Nutrient application was combined with suppression of mycorrhizal infection and with substrate structure modification in experimental patches. For each isolated root, we determined five dimensional characteristics and two topological parameters, including a newly introduced topological index (dichotomous branching index).

• Nonmycorrhizal L. campestris responded little to nutrient application, but strongly to benomyl application, in all characteristics measured. Mycorrhizal P. angustifolia produced the longest, most branched roots but exhibited limited sensitivity to nutrients and benomyl application. Strongly mycorrhizal P. lanceolata was the most sensitive to nutrient application, but showed little response to benomyl application. It was the only one among the species studied with root characteristics influenced (negatively) by increased production of total root biomass in the patches. Substrate structure influenced dimensional characteristics of Poa and Luzula roots, but not the topological indices.

• Results indicate different exploitation of soil microsites by L. campestris, P. angustifolia and P. lanceolata. Root topology seems to play a limited role in this process.


The soil environment of natural ecosystems is heterogeneous both in time and space, even on a small scale (Jackson & Caldwell, 1993; 1milauer, 1996; Ryel et al., 1996). Plant response to nutrient patchiness can include changes in biomass allocation, in uptake kinetics or in root morphology (Caldwell et al., 1992; Fitter, 1994; Fransen et al., 1998; Derner & Briske, 1999; Einsmann et al., 1999; Farley & Fitter, 1999; Ryser & Eek, 2000). Some of the competitive, dominant species do not respond to nutrient patches by the changes listed above, but their roots represent a large part of root biomass in patches because of their high growth rate (Campbell et al., 1991; Grime, 1994). Grime (1994) suggested that subordinate species forage by their roots more precisely than dominant species and that the dicotyledonous species forage more precisely than the grasses. According to Grime (1994), a trade-off exists between the scale (high for dominant plants) and the precision (high for subdominant plants) in resource foraging. Einsmann et al. (1999) did not confirm Grime’s (1994) prediction: they found scale and precision positively correlated in herbaceous species. Robinson & van Vuuren (1998) analysed published data of root response to nutrient patches for 27 wild plant species, which differed in growth rate (RGRmax) and in life form (grasses and forbs). They found that roots of fast-growing species proliferate into nutrient patches more precisely but only relative to uniformly nutrient-deficient controls, and that forbs proliferate more precisely than grasses.

The plasticity of root architecture was predicted to play an important role in the root response to soil heterogeneity (Fitter, 1994), but it has rarely been measured, especially as a response to nutrient patchiness (Fitter, 1994; Arredondo & Johnson, 1999). In an experiment where root response to increased nutrient availability was studied irrespective of spatial distribution of nutrients (Fitter et al., 1988), root topology was generally insensitive to changes in nutrient supply while the length of exterior and interior links (sensuFitter, 1991) and the specific root length changed with nutrient availability. Fitter and Stickland (1991) and Taub and Goldberg (1996) found the root topology of grasses less sensitive to changes in nutrient availability than in the dicotyledonous species.

Although the role of mycorrhizal fungi in nutrient acquisition by host plants, mainly in nutrient-poor soils (Read et al., 1976), is generally accepted (Koide, 1991; Wilcox, 1991), their contribution to nutrient acquisition from soil microsites has rarely been studied (St John et al., 1983a,b; Cui & Caldwell, 1996; Farley & Fitter, 1999). Tibbett (2000) ascribed a greater importance in nutrient patch exploitation to mycorrhizal mycelia than to plant root response. Moreover, root morphology seems to be directly influenced by mycorrhizal symbionts. Hetrick et al. (1988) and Hetrick (1991) suggested that plants highly dependent on mycorrhizal symbiosis reduce the metabolic cost of their roots by developing coarse root systems and that these changes in root architecture may be induced directly by mycorrhizal fungi. Fusconi et al. (2000) showed in Allium porrum that mycorrhizal fungi blocked root apical meristem activity. Nutrient requirements of plants change during the season depending on their phenological stage. These changes are reflected, for example, in seasonal dynamics of mycorrhizal symbiosis (Hetrick et al., 1988, 1994a, 1994b; Hartnett et al., 1993; Mullen & Schmidt, 1993; Lapointe & Molard, 1997).

Most experiments on root response to soil nutrient microsites have been performed under greenhouse conditions, usually with young plants and with no competition involved. The response of plant roots is often different when plants are grown individually and when under competition (Huber-Sannwald et al., 1998; Cahill & Casper, 1999; Fransen et al., 1999b; Robinson et al., 1999). Caldwell et al. (1991a, 1991b, 1996) found that root exploitation of nutrient patches by a plant of one species depends on species identity of neighbouring roots and on the size of competing plants. McConnaughay and Bazzaz (1992) found considerable differences among species in their sensitivity to space fragmentation by artificial root systems. Such a reduction in the growth of plants, which were constrained in deployment of their roots, was not detected at higher nutrient levels.

The aim of present study was to compare response of several distinct plant species from a grassland community to nutrient patchiness established under field conditions and to estimate the role of mycorrhizal symbiosis in this process. For this study, we chose three species occurring together in a nutrient-poor semi-natural grassland: one forb, Plantago lanceolata, and two graminoids, Luzula campestris and Poa angustifolia. They are dominant species on this site and differ in ecologically important characteristics (e.g. type of clonal growth, phenological separation in season and extent of mycorrhizal dependence). Attention was paid only to root morphological response; physiological plasticity was not studied.

The following questions were addressed in this study:

  • 1Do roots of the coexisting species respond similarly to a spatially localized increase in the nutrient availability?
  • 2Is their response modified by suppression of symbiotic mycorrhizal fungi?
  • 3Does substrate structure modify the morphological and architectural characteristics of roots growing into a newly created soil space?

Since the total root production in the experimental patches differs not as a result of the type of treatment, but also depends on the surrounding community composition (Šmilauerová, 2001), we asked an additional question:

  • 4Is the response of target species correlated with the growth response of neighbouring community (quantified by the total below-ground biomass in-grown into the patches)?

Materials and Methods

Research site

The research site is located near Zvíkov village (10 km east of České Budějovice, 48°59′ N, 14°36′ E, 500 m above sea level). The soil type is Cambisol, developed on paragneiss bedrock. The soil concentration of available macro-nutrients is low (per 100 g of soil dry wt): 0.22 mg NH4+, 0.06 mg NO3 and 0.38 mg of inorganic extractable phosphates. The vegetation represents an oligotrophic, traditionally managed meadow on a shallow valley slope. The meadow is cut once a year, usually in the middle of June. More information can be found in 1milauer and Šmilauerová (2000). The plant species nomenclature follows Rothmaler (1976).

Target species

We intended to compare grassland species differing in their mycorrhizal dependence and in their growth form. Therefore, three target species were chosen in this study: Luzula campestris (L.) DC, a rhizomatous nonmycorrhizal species with early spring activity, Poa angustifolia L., a rhizomatous mycorrhizal species flowering in June, and Plantago lanceolata L., a rosette forb with strong mycorrhizal colonization and with a long reproductive period from June to the end of summer.

Experimental design and procedure

Three separate experimental runs with an identical design were established during 1999, each one represented by a single experimental plot. Each of the runs started at that part of season when one of the target species flowered so that these species were in a comparable phenological stage (for characterization of these three experimental runs see Table 1).

Table 1.  The characteristics of the three experimental runs
 Target species Luzula campestrisPoa angustifoliaPlantago lanceolata
  • 1

    The codes for neighbour species are: AM, Achillea millefolium; CH, Cerastium holosteoides CP, Campanula patula; FR, Festuca rubra; HL, Holcus lanatus; LC, Luzula campestris; PA, Poa angustifolia; PL, Plantago lanceolata.

  • 2

    2 Average soil temperature measured at one reference point in the period of the respective experimental run.

  • 3

    3 Average soil water suction values measured at two reference points, in the period of the respective experimental run.

Sample period18 March−10 May 19991 April−25 May 19991 June−24 July 1999
Most frequent neighbour species1LC, PA, AM, PL, CHPA, CP, AM, FR, PLPL, LC, PA, FR, HL
Soil propertiesWell drained soil with poorlySoil with more developedAs for the run
developed H horizonH horizonwith L. campestris
Soil temperature (°C)2  9.3 10.3 20.8
Soil moisture (kPa)3−8.1−9.5−3.2

In each of these three experimental runs, there were seven replicate blocks. Each replicate block contained all 12 combinations of the experimental treatments (two levels of nutrients factor × two levels of fungicide factor × three types of substrate modification) randomly allocated to 3 × 4 sampling points arranged in a rectangular grid with span of 0.25 m. To establish a patch, soil from the soil profile was taken near an individual of the target species (situated close to one of the sampling points) with the aid of soil corer (diameter 4.5 cm, depth 10 cm). The soil was then sieved to remove roots and rhizomes (sieve mesh size 3 mm) and either put back into the hole or mixed with fine (diameter < 1 mm) or coarse (diameter 3–5 mm) sand in 1 : 1 ratio and then placed back into the hole. Nutrients in the form of phosphate, nitrate, and ammonia ions (using 0.2 g Na3PO4·12H2O, 0.24 g NaNO3, and 0.15 g NH4Cl per patch and application) and the fungicide benomyl (Bavistin, BASF, Ludwigshaften, Germany; 0.15 g Bavistin per patch) were applied in 0.1 l of water per patch. The control patches were treated with the same amount of water. The application of nutrients, fungicide, and water was repeated 3 wks after the first application.

Patches with surrounding soil and plants (soil columns with diameter of 11 cm and depth approx. 12 cm) were collected after 53 d of the experimental run. The samples were taken only from five or six replicate blocks in each experimental plot; the remaining patches were used for other purposes, including chemical analyses. From each patch, in-grown roots of one or more individuals of the target species were isolated using a needle and without water, tracking their connections to above-ground parts. Only the root parts growing within the patches were used for analyses of root characteristics. After isolation of the target plant roots all the remaining roots were separated from the patch substrate. Roots and rhizomes collected in the beginning and at the end of experiment were washed, dried (80°C, for 24 h) and weighed.

Separated roots of the target species were spread on a glass plate and their images scanned using a flatbed scanner with resolution of 600 DPI (Umax Astra 1220, UMAX Technologies Inc., Dallas, TX, USA, with transparency adapter). Root architecture, recorded in the image files, was evaluated using the rootarch software (P. 1milauer, unpublished). We used several morphological root characteristics for the statistical analyses: ELL, the average length of exterior links; ILL, the average length of interior links; µ, the magnitude of the root (number of root tips); TotL, the total length of the root, TotL : µ, the average root length per root tip and two topological indices log(pe) : log(µ); and dichotomous branching index (DBI) (see next paragraph). All the characteristics, except the last one, are described in Fitter (1991). Primary morphological characteristics are defined in Table 2. In statistical analyses, unbranched roots were considered to consist of a single exterior link.

Table 2.  Definitions of primary root morphological characteristics used in this paper
EL (exterior link)Terminal part of root between the root tip with meristem and the nearest branching point
IL (interior link)Root part joining other links (i.e. the part of root between any adjacent branches)
µ (magnitude)Number of exterior links (i.e. root tips) served by a root
TotL (total length of root)Sum of the lengths of all exterior and interior links of the root, expressed in millimetres
pe (total exterior path length)Sum of the number of links in all paths from any exterior link to the base of the root
max(pe) (maximal total exterior path length)Total exterior path length of imaginary root of given magnitude if fully ‘herringbone’-style branched
min(pe) (minimal total exterior path length) Total exterior path length of imaginary root of given magnitude if fully dichotomously branched
  • The newly introduced topological index, DBI, is calculated for a particular root system with the magnitude µ and the total exterior path length pe as

DBI = [pe − min(pe)]/[max(pe) − min(pe)]

This index shows the relative position of the actual total exterior path length value of a root between the reference values min(pe) and max(pe). Its values are therefore between 0 and 1, and so it is easier to estimate the position of the root on the scale between fully dichotomously and fully herringbone-style branched roots of the given magnitude. Moreover, this index seems to be scale-independent, which is not true for the other commonly used topological indices (e.g. pe : max(pe) or pe : E(pe), where E(pe) is the expected value of the parameter pe in the case of random branching).

The parameters analysed can be divided conceptually into two groups. One group (ILL, ELL, µ, TotL and TotL : µ) is referred to as dimensional parameters in the results and discussion, while the other group contains the topological indices log(pe) : log(µ) and DBI.

Supplementary measurements

To estimate soil moisture dynamics during the experiments, sensors for measuring the water saturation deficit of soil (by a portable digital metering device, based on soil resistance measurements, Watermark 30KTCD, Irrometer Co., Riverside, CA, USA) were placed in patches with one of the three applied substrate modifications near the experimental plots during the experimental runs. Two sensors were also placed into the soil profile among the three experimental runs for permanent measurements in unmanipulated soil during the whole season.

The effectiveness of benomyl application in the suppression of mycorrhizal colonization was evaluated on root samples from 145 soil patches with all treatment combinations, from all the experimental runs. Roots were stained using the modified Phillip & Hayman (1975) procedure, with Chlorazol Black E stain. Arbuscular mycorrhizal colonization was examined using a microscope (Olympus BX50), Olympus Optical Co., Tokyo, Japan) at magnification of ×400 and ×200. The percentage of root length colonized was then estimated for the whole sample at magnification of ×100 and ×45. Any root part with mycorrhizal structures (arbuscules, vesicles or hyphae apparently connected with arbuscule-bearing hyphae) was considered to be colonized by arbuscular mycorrhizal fungi.

Statistical analyses

For statistical analyses where the response variable was measured on individual roots, an analysis of variance including additional error level (corresponding to plant identity, below the error level of individual cores, at which the experimental treatments were applied) was used (nested anova). All response variables, except log(pe) : log(µ), were log-transformed to suppress their heteroscedasticity. The effect of the experimental blocks within each of the three runs was also modelled using a factor with a random effect. When analysing the effect of ingrown root biomass upon the root morphological properties, the same nested model approach was used, this time using a quantitative predictor. When using the two topological indices as response variables, only the roots with more than three root tips were included, as the two extreme topologies (the herringbone and dichotomous branching) cannot be distinguished for less branched systems.

When the amount of ingrown root biomass was used as a predictor, it was also log-transformed, because we expected its effect upon the root characteristics to be multiplicative (invoking a unit change in the response variable, with biomass increasing by a constant percentage amount). If the effect of root biomass was also significant, we tested its conditional effect, exhibited in addition to the effects of experimental treatment. A significant conditional effect can be then interpreted as an effect of root biomass amount, not explainable by the experimental manipulation.

When analysing the response of root characteristics to experimental factors or to amount of ingrown root biomass, an attempt was always made to build a common model for all three species, and we included the target species effect into the anova model. Further, we compared this model, where the effects of target species identity and of experimental factors were additive, with an alternative model where the treatment effects were nested within the host species effect. If the latter model was significantly better (as judged by an F-ratio test on the reduction in the residual sum of squares), the anova model was then fitted separately for each of the three target species, because the test outcome implied that the three species responded in a different way to the same combinations of experimental factors.

Analysis of the experimental treatment effects on the mycorrhizal colonization used a generalized linear model (GLM, McCullagh & Nelder, 1989) with an assumed Poisson distribution and logarithmic link function, because the subjective estimates of percentage colonization cannot be modelled with the assumption of binomial distribution. No additional error level was assumed here because we had just one root sample from each core and the replication occurred at the core level. The dependence of percentage of unbranched roots upon the experimental treatment was modelled using a GLM with an assumed binomial distribution and logit link function. To partly suppress the effect of overdispersion, the F-ratio based test in analysis of deviance was used (McCullagh & Nelder, 1989).

All the statistical models were fitted and tested using the s-plus for Windows 4.5 software (MathSoft, 1999).


The values of morphological characteristics of roots that grew into the patches, averaged over all the experimental treatments, are summarized for each of the target species in Table 3. Poa produced the longest roots, which had the largest proportion of dichotomous branching, and had numerous, very short exterior links. The ingrown roots of Luzula were the shortest among the three species, with branching restricted almost exclusively to the main axis (herringbone type of branching) and with their interior and exterior links slightly longer than in Poa. Plantago produced the smallest number of root tips per in-grown root, but both its exterior and interior links were the longest among the three species. The branching pattern of its roots was more dichotomous than for Luzula roots and its average root was twice as long as the average root of Luzula.

Table 3.  Medians, means and standard errors of the morphological characteristics of roots that grew into newly created patches, summarized for each species over all treatments 2
Characteristic1ILL (mm)ELL (mm)µTotL (mm)log(pe) : log(µ)DBIn3
  1. 1 ILL, average length of interior links; ELL, average length of exterior links; µ, magnitude of the root (number of root tips); TotL, total length of the root; log(pe) : log(µ) and DBI, two topological indices; DBI is dichotomous branching index, DBI = (pe − min(pe))/(max(pe) − min(pe)). For definitions of primary root morphological characteristics see Table 2. 2 Because many of the variables do not have symmetrical distribution, they are also characterized by the median estimates. 3n, Number of observations used to calculate the summary statistics.

Luzula campestris
Median1.201.78 27 791.810.96 
Mean2.145.15 742231.800.83218
± SE0.501.08  8.2 240.0030.02 
Poa angustifolia
± SE0.130.05 21.0 580.010.02 
Plantago lanceolata
Median3.805.90 161541.790.84 
Mean4.778.04 252391.770.74383
± SE0.430.86  1.8 150.0040.01 

The morphological characteristics of the ingrown roots responded to experimental manipulation in a qualitatively different way for the three species (P < 0.001 for ILL (F22,251= 2.481); P < 10−4 for ELL (F22,251 = 2.744); P < 0.001 for DBI (F22,251 = 2.506); P < 10−5 for TotL (F22,251 = 3.390); P < 10−4 for TotL : µ (F22,251 = 2.726), P < 10−7 for µ (F22,251= 4.100); P < 10−7 for log(pe) : log(µ) (F22,251 = 4.147)), therefore separate analyses were made for individual target species (experimental runs).

The effect of experimental treatments upon the morphological characteristics of the ingrown roots of individual species are summarized in Table 4 (for Luzula), Table 5 (for Poa) and Table 6 (for Plantago), and in Figs 1 and 2.

Table 4.  Response of morphological characteristics of Luzula campestris roots to experimental factors
Response variableILLELLµTotLTotL : µlog(pe) : log(µ)DBI
  1. anova model included a full interaction between the treatment effects, but the table shows all the main effects and only the interaction terms that were significant for at least one response variable (the third-order interaction term in this case). The estimate of Type I error probability is followed (in brackets) by the F-ratio statistic, on which it is based. The number of residual dfs is equal to 104 for log(Pe) : log(µ) and DBI, 106 for ILL, and 109 for the other response variables. There are further 97, 101 or 130 residual DFs at the ‘within-core’ error level. The test DF is equal to 1 for the N and B factors and 2 for the S effect and the N : B : S interaction term. See Table 3 for the definitions of the response variables.

Nutrients (N)nsnsnsnsnsns0.022
Fungicide benomyl (B)< 10−5< 10−5< 10−5< 10−5< 10−5< 10−4< 10−5
Substrate structure (S)nsns0.012ns0.019nsns
Interaction N : B : Sns0.017ns0.0380.045nsns
Table 5.  Response of morphological characteristics of Poa angustifolia roots to experimental factors
Response variableILLELLµTotLTotL : µlog(pe) : log(µ)DBI
  1. anova model included a full interaction between the treatment effects, but the table shows all the main effects and one of the second-order interaction terms, as the other interaction terms were nonsignificant for any of the response variables. The estimate of Type I error probability is followed (in brackets) by F-ratio statistic, on which it is based. The number of residual dfs is 45 for all the response variables. There are further 140 residual dfs at the ‘within-core’ error level. The test df is equal to 1 for the N and B factors and 2 for the S effect and the B : S interaction term. See Table 3 for the definitions of the response variables.

Nutrients (N)ns0.038nsnsnsnsns
Fungicide benomyl (B)0.034nsnsns0.024nsns
Substrate structure (S)ns0.024nsnsnsnsns
Interaction B : Sns0.005nsnsnsnsns
Table 6.  Response of morphological characteristics of Plantago lanceolata roots to experimental factors
Response variableILLELLµTotLTotL : µlog(pe) : log(µ)DBI
  1. anova model included a full interaction between the treatment effects, but the table shows only the main effects, as the interaction terms were nonsignificant for all the response variables. The estimate of Type I error probability is followed (in brackets) by F-ratio statistic, on which it is based. The number of residual DF is equal to 102 for log(Pe)/log(µ) and DBI, to 104 for ILL, and to 106 for the other response variables. There are further 265, 277, or 294 residual DFs at the ‘within-core’ error level. The test DF is equal to 1 for the N and B factors and 2 for the S factor. See Table 3 for the abbreviations of the names of response variables.

Nutrients (N)< 10−5< 10−3nsns< 10−5nsns
Fungicide benomyl (B)nsnsnsnsns0.044ns
Substrate structure (S)nsnsnsnsnsnsns
Figure 1.

Effect of nutrients application on the morphological characteristics of roots that grew into the experimental patches. Averages are shown by vertical bars separately for the three species studied (empty bars for patches without supplementary nutrients; filled bars for patches with addition of nutrients). Vertical lines indicate 95% confidence intervals. The significance of differences in respect to nutrients application can be found in Tables 4, 5 and 6 for, respectively, Luzula campestris (LC), Poa angustifolia (PA) and Plantago lanceolata (PL). The averages for the two levels of the nutrients treatment are taken over all levels of the other two experimental factors (fungicide and substrate structure). ILL, the average length of interior links; ELL, the average length of exterior links; µ, the magnitude of the root; TotL, the total length of the root; log(Pe) : log(µ) and DBI (dichotomous branching index) are two topological indices. For definitions of primary root morphological characteristics see Table 2.

Figure 2.

Effect of fungicide application on the morphological characteristics of roots that grew into the experimental patches. Averages are shown by vertical bars separately for the three species studied (empty bars for patches without added fungicide; filled bars for patches with fungicide application). Vertical lines indicate 95% confidence intervals. The significance of differences in respect to fungicide application can be found in Tables 4, 5 and 6 for, respectively, Luzula campestris (LC), Poa angustifolia (PA) and Plantago lanceolata (PL). The averages for the two levels of the fungicide treatment are taken over all levels of the other two experimental factors (nutrients and substrate structure). ILL, the average length of interior links; ELL, the average length of exterior links; µ, the magnitude of the root; TotL, the total length of the root; log(pe) : log(µ) and DBI (dichotomous branching index) are two topological indices. For definitions of primary root morphological characteristics see Table 2.

Plantago roots responded to increased availability of nutrients in most of their dimensional parameters, but no change in root topology was found. In the patches with added nutrients, Plantago had shorter exterior as well as interior links and the average root length per root tip decreased with the application of nutrients. Poa responded to nutrient enrichment only by production of slightly shorter exterior links. The topology (as expressed by DBI) was influenced by nutrient application only in Luzula roots, indicating decreased dichotomy of branching.

Application of benomyl affected all root characteristics of Luzula, while both Poa and Plantago responded only slightly. In patches treated with benomyl, Luzula produced shorter, less branched roots with longer interior and exterior links. Number of root tips (µ) decreased by a greater extent than the total root length, as indicated by the significant increase of average root length per root tip. Both topological indices indicate a decreased extent of dichotomous branching in roots that grew into the benomyl-treated patches. Poa had longer interior links and greater root length per root tip in patches treated with benomyl, whereas Plantago responded to benomyl application only by slightly changed branching pattern (more dichotomously branched roots), but this was indicated only by the log(pe) : log(µ) index. No interaction was found between the effects of nutrient and benomyl application.

Substrate modification influenced some dimensional characteristics of Poa and Luzula roots, but not their topological indices. Luzula had more root tips (higher µ) and shorter root length per root tip in patches with a mixture of soil and coarse sand; the response to fine sand was in the opposite direction, and values of both root characteristics from patches with unamended sieved soil were between these two extremes. The average exterior link length of Poa roots was greater in patches with soil mixed with sand than in the sieved soil (the longest links were in patches with the fine sand fraction). The effect of benomyl application upon the average exterior link length of Poa roots was significantly influenced by manipulation of substrate structure in the patches.

To further explore the response of Luzula roots to fungicide application, the overall tendency of roots of the target species to produce poorly branched roots (with magnitude µ < 7) in patches with different treatments was analysed, using a generalized linear model. For Poa, only three roots were poorly branched, each from a patch with a different treatment combination, so this species was not included in this analysis. Luzula produced a significantly larger proportion of poorly branched roots in patches with benomyl application (F1,64 = 5.52, P = 0.022), while for Plantago this proportion was not significantly different among the different treatments (F1,54 = 0.09, ns).

To compare behaviour of roots of the target species with the growth response of neighbouring community, the relation between the root characteristics of the three species and the total root biomass proliferated into the patches was analysed. A model with an additive relationship between the effects of the target species and of the root biomass was compared with another model, where the biomass effect was nested within the species effect. This comparison showed that the effect of root biomass differed among the target species (P < 0.001 for ILL (F2,803 = 17.25), ELL (F2,854 = 8.41), µ (F2,854 = 14.09) and TotL : µ (F2,854 = 18.72); P = 0.002 for log(pe) : log(µ) (F2,783 = 6.16); P = 0.0011 for DBI (F2,783 = 6.89); and P = 0.0016 for TotL (F2,854 = 6.50)). Table 7 summarizes results of separate analyses for the individual target species. After accounting for effect of experimental manipulation, the effect of total root biomass remained significant only for Plantago, in majority of its root characteristics.

Table 7.  Regression models of the dependence of measured root characteristics on the total root biomass proliferated into the experimental patches during the experiment
 ILLELLµTotLTotL : µlog(Pe) : log(µ)DBI
  1. The regression models were fitted separately for each of the three target species. Rows marked with asterisk (*) correspond to models where the effects of experimental manipulation were also included. ▴ indicates an increase in the values of oot morphological characteristic with the increasing total root biomass; ▾ indicates a decrease in root morphological characteristic values. na, not analysed; ns, not significant, for other abbreviations see Table 3. The estimate of the Type I error probability is followed (in brackets) by the F-ratio statistic, on which it is based.

Luzula▴ 0.03ns▾ 0.006▾ 0.009▴ 0.023▴ 0.01▴ 0.005
 (0.57) (1.67)(2.86)(0.002)(0.78)(0.52)
Poans▾ 0.02nsnsnsnsns
Plantago▾ < 10−5▾ < 10−3)▴ 0.008ns▾ < 10−5▾ 0.02ns
Plantago*▾ 0.015▾ 0.044nsna▾ 0.007▾ 0.05na

We also checked the effectiveness of benomyl application in decreasing the mycorrhizal colonization. No mycorrhizal colonization was found in Luzula roots from any treatment. Nonmycorrhizal fungi were also very rare in these roots. Poa and Plantago roots differed in the percentage of mycorrhizal colonization (F1,102 = 35.37, P < 10−8): Plantago roots in control patches had mycorrhizal colonization in more than half of their length (51.2%, SE = 3.06), while Poa roots had less than 10% of their length colonized (9.6%, SE = 2.67). The reduction of mycorrhizal development by fungicide was significant only for Plantago roots (F1,49 = 38.17, P < 10−5; Fig. 3), but colonization intensity was also substantially reduced in Poa roots.

Figure 3.

Percentage of root length with developed arbuscular mycorrhizal symbiosis in roots of two target species (Plantago lanceolata and Poa angustifolia) in individual combinations of two experimental factors (application of nutrients (N) and application of fungicide (B)). Treatment 0 corresponds to cores where neither nutrients nor fungicide were added; cores with treatment NB were supplemented with both nutrients and fungicide solution. The measurements are averaged over all three levels of the third experimental factor (substrate structure) because the mycorrhizal infection level was not significantly different among the substrate types.


The results show that the three grassland plant species studied differed significantly in their response to experimentally induced soil heterogeneity.

Luzula, with its slow growth, short links and the most ‘herringbone’ topology (see Table 3) is a typical species of nutrient-poor sites (Fitter et al., 1991). There are several possible explanations of its low sensitivity to nutrient application. The first explanation is that the duration of experiment was too short for a significant response by a slow-growing species. Nevertheless, the exploitation of nutrients from patches by microorganisms and plants can be quite rapid (Fransen et al., 1998; Hodge et al., 1998) and, therefore, the response to a newly established patch has to be sufficiently speedy so that expenses do not exceed profit. Small nonsignificant changes of all measured characteristics indicate some ‘hesitation’ to branch. This can represent an economic strategy, when in the nutrient-enriched patches the amount of nutrients sufficient for species adapted to nutrient-poor sites can be acquired by the less-branched roots. The more ‘herringbone’ systems are considered to be more efficient in nutrient acquisition but more expensive to construct and maintain (Fitter, 1991). It is also possible that Luzula roots did not change their growth when encountering a nutrient-rich patch, but changed their uptake rate per root length (Caldwell, 1994). The low growth response of Luzula roots can be also partly influenced by the lower soil temperature at the beginning of this experimental run. Note, however, that the average soil temperature through this experimental run was similar to the average for the experimental run with Poa (Table 1).

Poa angustifolia was able to produce the longest roots in the newly created patches with most dichotomous branching and very short links (see Table 3). The ability of this species to create very fine, amply branched roots in vacant substrate patches was only slightly influenced by nutrients or fungicide application. It seems that Poa roots grow so fast that they are able to occupy the empty soil space in a short time and to acquire nutrients without any essential change of root morphological properties. The low topological plasticity of Poa roots found in this study agrees with the results of Taub and Goldberg (1996) for grasses, although they found grasses to have a maximum herringbone topology. Perhaps the seedling roots behave differently from the proliferating roots of adult individuals.

Plantago had the least branched roots (with a typical magnitude value of 16), but with the longest exterior and interior links. The growth rate of its roots was slower – if expressed by the total root length – than the growth rate of Poa roots. Plantago responded to the addition of nutrients by production of more compact roots (with shorter links and lower average root length per single root tip), which agrees with results of the Fitter et al. (1988) study but not with those in Fitter (1994), where exterior links were longer in treatments with higher nutrient availability. Glimskar (2000) also found that a grassland forb, Polygala vulgaris, responded to low nitrogen supply by markedly increasing the length of root links. The increased compactness of Plantago roots from nutrient-rich patches observed in our experiment may contribute to increased competition among roots of the same plant and its profitability would depend on the properties of the patch (Fitter, 1991, 1994).

Insensitivity of the morphological properties of Luzula roots to nutrient addition strongly contrasts with its extensive negative response to the application of fungicide. Studies dealing with the plant response to fungicide application have measured the effect of fungicide on roots as the effect on the root biomass (Borowicz, 1993; Newsham et al., 1994) or on the root length per individual plant (Carey et al., 1992; Sukarno et al., 1993), but not on the root morphology. Response of nonmycorrhizal plants to benomyl application has been mentioned only by Fitter and Nichols (1988) for Sinapis alba and by Borowicz (1993) for Brassica napus. No effect of this fungicide on P inflow (Fitter & Nichols, 1988) or on root mass (Borowicz, 1993) was found. Several authors studied the influence of benomyl on other soil microorganisms and some metabolic processes. In addition to the suppression of nonmycorrhizal fungi (Carey et al., 1992; West et al., 1993), interference with the nitrogen cycle was found (Chen et al., 1995; Cademenun & Berch, 1997). Neither Fitter and Nichols, 1988) nor Merryweather and Fitter (1996) found any substantial increase in soil phosphate concentration as a consequence of benomyl application. The whole community response to benomyl application expressed by total root biomass grown into the patches was positive (Šmilauerová, 2001), and this can be explained either by suppression of some antagonistic soil fungi or by a short-term flush of nutrients from the components of soil biota adversely affected by the fungicide. Response of Luzula roots to benomyl application was, after removal of effects of experimental manipulation, independent of the total root biomass grown into the patches (Table 7), and this suggests that the depression of growth of Luzula roots was not caused by increased competition with the roots of other plant species. Because the response of Luzula differed from the response of the whole community, at least in its tendency not to proliferate into benomyl-treated patches, we consider the benomyl effect upon Luzula to be specific for this species. Although the root morphological response to benomyl application was much more pronounced than the response to nutrient enrichment, the direction of their effects was identical for each of the characteristics measured (Figs 1 and 2). It is therefore possible that ‘conservative’Luzula, adapted to nutrient-poor soils, is more sensitive to fluctuation of environmental characteristics exceeding the usual conditions. Encountering a soil microsite with increased nutrient concentration slightly exceeding the standard variability under natural conditions could cause just a small delay in root growth, whereas a rapid decline in microorganismal activity after benomyl application or occurrence of some byproducts of benomyl decomposition could be so far from the standard conditions, that a strict negative response was invoked. Alternatively, we may speculate about existence of different group of organisms with a mutualistic relationship with Luzula roots, which was suppressed by fungicide application. Nevertheless, the negative response of Luzula to benomyl had probably just a short duration because the long-term restriction of root growth and branching would lead to a suppression of Luzula by more competitive species. However, such effect was not recorded in a long-term experiment performed on this site (1milauer & Šmilauerová, 2000; unpublished results for Luzula abundance).

Although the growth of Poa roots in new patches was very rapid, the experiment duration seems to be too short for the development of mycorrhizal symbiosis to the extent commonly seen in Poa roots at this site (average of colonized root length is 25%; unpublished). For this reason, the response of Poa roots to fungicide application was probably not a matter of mycorrhizal development restriction (in fact, mycorrhizal development was not significantly suppressed by fungicide application), but rather a side-effect of the fungicide application. Wilson and Hartnet (1997, 1998) found that the cool-season grasses from tallgrass prairie, including P. pratensis, are insensitive to mycorrhizal colonization or even benefit from reduction of mycorrhizal root colonization by benomyl. This indicates facultative mycotrophy of these grass species, although their mycorrhizal status in North American prairie can differ from that in European grasslands. The fact that the mycorrhizal colonization in Poa roots was suppressed more effectively by nutrient addition then by fungicide application (see Fig. 3), supports the idea of facultative mycotrophy of this grass species.

Plantago lanceolata has been used in numerous experiments as a typical mycorrhizal host species possessing a broad spectrum of mycorrhizal fungal taxa (Sanders & Fitter, 1991; Bever et al., 1996; Gange et al., 1999). The low response of Plantago roots to suppression of mycorrhizal colonization in present experiment is therefore surprising. Its roots at this study site have well-developed mycorrhizal structures (average colonization is 85% of root length; unpublished) and the mycorrhizal development was successfully restricted by benomyl in this experiment (Fig. 3) as well as in another, long-term experiment (1milauer & Šmilauerová, 2000). It is possible that in Plantago roots, the morphological changes induced by mycorrhizal fungi are not as marked as in some other species (Hetrick et al., 1988; Hetrick, 1991; Fusconi et al., 2000).

Substrate structure modification had only a slight effect and only on some of the dimensional parameters of Poa and Luzula roots. The greater magnitude of Luzula roots and their shorter average length per single root tip in patches with coarse sand mixture might be a result of higher branching frequency caused by contact of growing root tips with sand particles of sufficient size. Nevertheless, no similar response was found either in fine roots of Poa or in thicker roots of Plantago. Poa had longer exterior links in patches with sand. The response was opposite to the response of this root parameter to nutrient addition. It can be explained by a higher sensitivity of this parameter in Poa to changes in nutrient availability in patches combined with a dilution of soil nutrients in the mixture with sand. It is surprising that Plantago, with its highest sensitivity to nutrient addition, did not respond at all to the substrate structure modifications.

On the community level, the total ingrown below-ground biomass was 1.6 times higher in patches with added nutrients and 1.14 times higher in patches treated with benomyl, compared with the control patches (Šmilauerová, 2001). None of the three species studied responded to nutrient addition by an increased length of individual ingrown roots. Therefore, we should ask which root parameter changed so that the total below-ground biomass in treated patches increased. One possibility would be an increase of the diameter of proliferating roots but this is contradicted by most of the recent findings (Hetrick et al., 1988; Bilbrough & Caldwell, 1995; Hodge et al., 1998; Arredondo & Johnson, 1999). These studies found a significant increase in specific root length (SRL) in nutrient patches or in roots without mycorrhizal infection, while Fransen et al. (1999a) found a significant increase of root biomass in nutrient patches without any change in specific root length. The second root characteristic that could be responsible for the increase in total root biomass in the patches is the number of proliferating roots. As far as we know, no-one noted this characteristic in studies investigating root proliferation into nutrient patches or into soil microsites with restricted development of mycorrhizal symbiosis. A third possibility is that the total biomass increase was caused primarily by the other grassland species, which behave differently from the three species we have studied.

For root characteristics of Luzula (as well as for ELL in Poa and for magnitude in Plantago), it is not possible to separate the effect of total root biomass in the patches upon the measured root characteristics from the effect of experimental treatment (see Table 7). Plantago was the only one among the species studied that responded significantly to differences in total root biomass after accounting for the treatment effects. Its response to increasing total root biomass in patches (shorter links and shorter root length per one root tip) was similar to its response to nutrient addition, but the set of significantly modified root characteristics was extended by an increased dichotomy of branching. This suggests that this response can be a result of competitive suppression, rather than an economical tactic.

The adopted experimental design provides certain interpretational difficulties. The three experimental runs for the tree target species are temporally separated according to the flowering time of individual species. The flowering time represents a period of high physiological activity (as the start of energy-demanding development of seeds) and provides the best-defined reference point for seasonally differentiated activity of individual species populations. These shifts in seasonal activity must be taken into account particularly with species such as L. campestris, which have a major part of their reproductive cycle concentrated in the early parts of growing season. However, the different timing of experimental runs for individual species brings inevitable confounding of interspecies differences with the effects of differences in abiotic as well as biotic conditions. Despite this limitation, we still prefer this design, as it better shows the species studied in the context of their usual growing conditions.

In summary, we found considerable differences in root response of three grassland species to nutrient microsites and restriction of mycorrhizal colonization in the patches. An important conclusion from our study is that the root response to increased availability of nutrients was independent of the extent of mycorrhizal symbiosis.

The dominant grass species Poa angustifolia exhibited a high growth rate of roots entering the vacant soil patches, irrespective of the concentration of available nutrients or of the reduction of mycorrhizal activity. This might suggest a highly competitive, imprecise foraging in the soil space. The roots of L. campestris entered the available patch space very slowly and their response to increased availability of nutrients in the patch was negligible (slightly reduced root branching). This might be a typical response of a stress tolerator that is not well prepared to deal with such short-term opportunities. Luzula roots had a profound, but not yet understood response to fungicide presence. Roots of P. lanceolata responded to enriched soil patches by producing more densely branched root systems, suggesting a more precise targeting of local resource patches. Our results indicate higher sensitivity of Plantago roots to competitive effects of other species entering the patches. This might illustrate one of the mechanisms (additional to the increased competition for light) by which the populations of many forbs are eliminated from nutrient-enriched grasslands by the more competitive grasses.


We thank our technician Blanka DiviSová for her help with the sample processing and Alastair H. Fitter, Jan 1. LepS, Sylvie Pechá4ková, and two anonymous reviewers for their useful comments on the manuscript. The project was funded from the research grants no. 206/98/0047 and 206/99/0889 of the Grant Agency of the Czech Republic.