Species selection and plant material
To investigate the growth consequences of contrasting root behaviours in dynamic heterogeneous conditions, we build upon the knowledge available on root foraging traits in two co-occurring perennial grassland species of European hay meadows: Festuca rubra L. and Anthoxanthum odoratum L. F. rubra is a more precise forager in terms of root placement than A. odoratum, reflected in values of selective root placement of 3.5 and 2.4 (roots in rich/roots in homogeneous soil), respectively (Fransen et al 1999). A. odoratum is thought to express a higher physiological plasticity than F. rubra, measured as 15N uptake per unit root biomass per day, being 2.2 and 1.4 mg g-1 day-1, respectively (Fransen et al. 1998, 1999, 2001). Often, F. rubra achieves up to two times higher root densities than A. odoratum (Fransen et al. 2001; Mommer et al. 2010), but not always (e.g. Fransen et al. 1998). The species have comparable relative growth rates (Grime and Hunt 1975; Fransen et al. 1998).
Seeds of the two grass species were collected in the river forelands of the river Rhine, near Nijmegen (The Netherlands) and germinated for 10 days on moistened filter paper in Petri dishes. The seedlings were transplanted to small pots (diameter 5 cm) filled with a vermiculite:sand (1:1) mixture and grown for 4 weeks. The seedlings were watered with ¼ Hoagland solution three times a week.
Competition treatments and density series
After 4 weeks, the plants were transplanted to the actual experimental units, being bigger pots (diameter 23 cm, depth 16 cm) filled with the same vermiculite:sand (1:1) mixture. Within a pot, four plants were planted in a square, 9 cm apart.
The nutrient distribution treatments (stable, switch; see explanation below) were applied to monocultures and mixtures of the two grass species. Competitive ability was measured by comparing shoot growth in mixture to shoot growth in monoculture (de Wit 1960; Fransen et al. 2001). Every combination of nutrient treatment (stable or switch) × competition treatment (mono of A. odoratum or F. rubra or their mix) was replicated 18–22 times, resulting into 115 pots with plants to be harvested at the end of the experiment. Halfway through the experiment, at the moment of the switch in nutrient supply (see below), 60 pots were harvested.
To investigate whether competition occurred in our four plants per pot setup, we included a density series with two, four and eight plants of both species; the number of replicates was eight. If the plants in our four plants per pot actually competed strongly, total biomass per pot would be similar for four and eight plants per pot (Fransen et al. 2001). The density series only received the so-called stable treatment. Significant competition occurred in the experiment where a density of four plants was used, since final yield (maximum shoot biomass 10.6 and 8.8 g per pot for A. odoratum and F. rubra, respectively) in a density series had already levelled off between two and four plants per pot.
Nutrient heterogeneity treatment
We explicitly aimed for an experimental setup with multiple nutrient patches that were predictable to different degrees. At first, all plants were grown in pots with a heterogeneous nutrient application for 5 weeks. Thereafter, in half the pots, the nutrient-rich patches were continued as before (referred to as the stable treatment). In the other half of the pots, the nutrient status of some of the patches changed position (referred to as the switch treatment), thus confronting the plants with root investments at the ‘wrong’ location. Plants were then grown for another 4 weeks.
The heterogeneous treatment consisted of three types of patches that were created at the start of the experiment (Fig. 1). First, in the centre of the pot, a stable nutrient-rich patch was created. This nutrient patch, containing half of the nutrients, was permanently available. Second, two nutrient-rich patches were created at the outer side of the pot (i.e. peripheral patches on the horizontal axis in Fig. 1). These patches continued to be nutrient–rich in the stable treatment, but in the switch treatment they changed into nutrient-poor patches after 5 weeks. Third, two nutrient-poor sites were also created at the outer side of the pot (i.e. peripheral patches on the vertical axis in Fig. 1). These patches continued to be nutrient–poor in the stable treatment (i.e. background soil nutrient status), but in the switch treatment they changed into nutrient-rich patches after 5 weeks. The combination between switch treatment and position resulted in four different peripheral patch types (rich–rich, poor–poor, rich–poor, poor–rich). The nutrient-rich patch in the centre of the pot received an equal amount of nutrient solution as the two nutrient-rich peripheral patches together, i.e. the stable centre patch was twice as large as each of the peripheral patches. The distance of each plant to the three nearest patches was the same, so the chance that the plants encountered one of the three patches was the same. The switch treatments differed in patch distribution, but received the same total amount of nutrients at any time.
The patches were created using a dripping system with nutrient solution three times a week (cf. Campbell and Grime 1989; Jansen et al. 2006; Visser et al. 2008). The nutrient-rich patches received 15 ml 1/2 strength Hoagland’s solution [3 mM KNO3, 2 mM Ca(NO3)2·4H2O, 1 mM NH4H2PO4, 0.5 mM MgSO4·7H2O, Fe-EDTA and micronutrients] and the nutrient-poor patches received 15 ml of a very poor background nutrient solution [1/64 strength Hoagland’s, which is lower in nitrate concentration than the Dutch standards for tap water; 0.09 mM KNO3, 0.06 mM Ca(NO3)2·4H2O, 0.03 mM NH4H2PO4, 0.015 mM MgSO4·7H2O, Fe-EDTA and micronutrients]. To prevent leaching of nutrients from one nutrient hotspot in another nutrient patch, nutrient-poor background solution (1/64 strength Hoagland’s) was dripped in between the different nutrient-rich patches (cf. Jansen et al. 2006; see Fig. 1).
To quantify the nutrient concentration of the patches, soil nitrate concentration was analysed at different spots in separate pots with and without plants. Soil cores (5–10 g, diameter 1 cm, 0–10 cm deep) were taken 2 h and 2 days after the nutrient solution had been applied. The soil sample was diluted in 100 ml demineralised water and shaken for 1 h by 120g to dissolve the nitrate in the water. The mixture was filtered and stored in plastic pots in a freezer. The samples were later analysed colourimetrically with a continuous flow analyser for nitrate (Traacs 800+; Brann and Luebbe, Germany). Nutrient-rich patches contained 0.870 ± 0.089 μmol NO
g−1 soil, nutrient-poor patches 0.230 ± 0.025 μmol NO
g−1 soil 2 h after dripping. Nutrient-rich patches contained thus more than 3.5 times more nitrate than nutrient-poor patches, indicating that the dripping method with the nutrient solution in syringes resulted in a heterogeneous soil. In pots where plants were present, all available nutrients were taken up within 2 days (when nitrate concentrations had dropped to 0.031 ± 0.010 μmol NO
g−1 soil) with no significant differences anymore between the patch types. The term ‘dynamic’ used with regard to the switch treatment refers to the complete switch in peripheral position of the nutrient solution halfway through the experiment. However, one may argue that even the permanent patches showed some degree of nutrient dynamics as the nutrients were consumed within 2 days and then reappeared again.
In order to avoid water deficit, all pots were watered once a week with 300 ml of the very nutrient-poor nutrition solution (1/64 Hoagland’s). This background solution was applied 2 h before nutrient addition. The applied nutrient levels were limiting since N contents of F. rubra and A. odoratum shoots (living and dead biomass) at final harvest were, averaged over all treatments, only 1.3 and 1.1% of dry biomass, which is well below standard values of 2–5% N of well-fed plants (Marschner 1995).
The position of the pots in the greenhouse was arranged randomly to homogenise growth conditions among the replicated plants. The experiment was performed in the heated greenhouse of the Radboud University Nijmegen (20°C, PPFD was between 200 and 550 μmol m−2 s−1, 16 h/8 h) from February to April 2006.
In summary, we performed a competition experiment with two grass species, growing in monoculture and mixture. All nutrients were applied heterogeneously in three different types of patches in two different ways: stable nutrient heterogeneity or partly dynamic. In order to determine nutrient uptake from the different patch types, we applied 15N pulses in all the patch types separately. Replicates for 15N application in the three patch types were 6, but root responses could be determined from all pots, leading to a replicate number of 18 (+3–4 for 15N background concentrations).
All destructive measurements were performed 5 weeks (mid harvest) and 9 weeks (final harvest) after the treatment had started. Initial plant biomass was also determined at the start of the experiment.
Shoot biomass, root biomass and root length
At mid- and final harvest, shoots were clipped at the root base, dried at 70°C for 48 h and weighed to determine biomass. At the same time, from the same pots, soil cores of 16 cm depth were taken with a device that allowed coring at five locations in the pot simultaneously. The five locations were the centre patch (diameter = 56 mm), the two peripheral nutrient-rich patches (diameter = 40 mm) and the two peripheral nutrient-poor patches (diameter = 40 mm). Note that the position of nutrient-rich versus poor patches in the pot depended on the treatment (see Fig. 1). The two pairs of peripheral patches (with similar nutrient regime) were summed, resulting in three different root sections: one from the nutrient-rich patch in the centre, one from the peripheral rich patches, and one from the peripheral poor patches, all with the same soil volume. The roots in the cores were carefully washed from the soil. It was impossible to visually disentangle the roots from the mixtures by species.
Sub-samples of the roots from the different patches were taken to determine specific root length. These root sub-samples were conserved in 0.01% HgCl2 solution at 4°C for a few weeks. Before scanning (at 600 dpi; Epson expression 10000 XL; Regent Instruments, Canada), the roots were coloured with a Neutral Red solution (0.5 g L−1) to improve contrast. Scans were analysed with WinrhizoTM software (Regent Instruments, Canada). Afterwards, the scanned root samples were dried at 70°C for 48 h in order to calculate specific root length (m g−1). Selective root placement of the two species and the mixture was expressed as root length density (m dm−3) in the target patch divided by the root length density of the continuously poor nutrient patch. Root length density was calculated from the root biomass from a patch, specific root length and the soil volume of the respective patch. The remaining roots in the pot were also washed in order to determine total root mass.
15N and total N uptake
We investigated nutrient uptake in the different patch types at both mid and final harvest by labelling with 15N either to (1) the central nutrient-rich patch, (2) the two peripheral nutrient-rich patches or (3) the two peripheral nutrient-poor patches. The three 15N applications were divided over the total of 18 replicates, i.e. 6 replicates per position of 15N application. Three to four replicates were used to determine natural background concentrations of 15N. We added 15N to the nutrient solution in each of the two syringes per patch type, during the last three dripping sessions (i.e. 5, 3 and 1 days) before the final harvest.
For each 15N application, 1 ml demineralised water containing 1 mg 15N (supplied as K15NO3; = 9 μmol ~ nitrate equivalent of 15 ml of 1/32 Hoagland’s solution) was added to the ‘normal’ gift of 15 ml of ½ Hoagland’s solution. In the case of the nutrient poor syringes, 14N-nitrate from the 1/64 nutrient solution was replaced for a 15N equivalent of 1/32 Hoagland’s solution (9 μmol 15N in 16 ml demineralised water). Thus, while differences in total nutrient concentration of the patches remained, every patch received an equal amount of 15N. Total amount of K15NO3 supplies was 6 mg, leading to a total amount of 0.84 mg 15N given per pot, which means that on average 0.21 mg 15N was available per plant (planting density = 4). On average, 50% of the supplied 15N to each pot was taken up by the shoots.
For the analysis of both 15N and total N, all shoot material of single plants was dried and pulverized and 2–2,5 mg of this shoot material was used in a nitrogen analyzer (EA 1110, Carlo Erba; Thermo Electron, Milan, Italy) in combination with a mass spectrometer (DeltaPlus; Thermo Finnigan, Bremen, Germany). The natural available background of 15N concentration, determined on every species × treatment combination (average background 15N concentrations for all species × treatment combinations 0.367%; no significant differences between species and treatments) was subtracted from the 15N concentrations of the labelled plants in order to determine excess 15N concentrations.
The effects of intra- versus inter-specific competition and heterogeneous nutrient dynamics on shoot biomass and total N accumulation were assessed using univariate ANOVA (SPSS 15.0; GLM), with species (A. odoratum vs. F. rubra), planting (monoculture vs. mixture) and treatment (stable vs. switching dripping regime) as fixed factors. These analyses were performed on species averages per pot (n = 18–21).
In statistical analyses using root length density the factors ‘species’ and ‘planting’ were combined in one fixed factor (referred to as ‘species combination’) with three levels: A. odoratum, F. rubra and ‘mixture’ (i.e. the response of A. odoratum and F. rubra as a whole), since the roots of the two species in mixture could not be disentangled quantitatively.
Analyses regarding root length density and 15N uptake included, next to effects of species and planting, respectively, effects of patch type. Since we were not interested in effects of patch size per se, statistical analyses for root length density and 15N uptake were run separately for the centre patches and peripheral patches, although root and nutrient uptake responses to centre and peripheral patches will not have been completely independent. For the centre patches in both stable and switch treatment, nutrient status was always nutrient-rich, since the switching only affected the peripheral patches. For the peripheral patches, the switch treatment and the position of the patches resulted into 4 different patch types with a different nutrient status in time (rich–rich, poor–poor, rich–poor and poor–rich), of which two were within a pot and not completely independent either. In the centre patches, the nutrient status was thus not affected by the switch treatment, whereas the nutrient status was affected by the switch treatment in the peripheral patches. This was an additional reason to run separate statistical analyses for the centre patches and peripheral patches. In order to test the effect of interdependence in the peripheral patch types, we also performed a statistical analysis with patch position nested within treatment (results not shown). As this yielded basically similar results as the factor patch type with four levels, we chose to present the most simple type of analysis, i.e. four levels of peripheral patch types.
For the monocultures of A. odoratum and F. rubra, analyses of 15N uptake were performed with root length density as a covariate factor. This covariate factor was not significant as root length density was driven by the factor patch type. Ln transformations of the data were performed when needed to meet the assumptions of ANOVA.