Foliar and soil δ15N values reveal increased nitrogen partitioning among species in diverse grassland communities

Authors

  • MARLÉN GUBSCH,

    Corresponding author
    1. Institute of Plant, Animal and Agroecosystem Sciences, ETH Zurich, Universitaetsstrasse 2, CH-8092 Zurich, Switzerland
      M. Gubsch. e-mail: mgubsch@gmail.com
    Search for more papers by this author
  • CHRISTIANE ROSCHER,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    2. UFZ, Helmholtz Centre for Environmental Research, Department of Community Ecology, Theodor-Lieser-Strasse 4, D-06120 Halle, Germany
    Search for more papers by this author
  • GERD GLEIXNER,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    Search for more papers by this author
  • MAIKE HABEKOST,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    Search for more papers by this author
  • ANNETT LIPOWSKY,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    2. Institute of Evolutionary Biology and Environmental Studies, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland
    Search for more papers by this author
  • BERNHARD SCHMID,

    1. Institute of Evolutionary Biology and Environmental Studies, University of Zurich, Winterthurerstrasse 190, CH-8057 Zurich, Switzerland
    Search for more papers by this author
  • ERNST-DETLEF SCHULZE,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    Search for more papers by this author
  • SIBYLLE STEINBEISS,

    1. Max Planck Institute for Biogeochemistry, POB 100164, D-07701 Jena, Germany
    2. Institute of Groundwater Ecology, Helmholtz Zentrum Munchen, German Research Centre for Environmental Health GmbH, Ingolstaedter Landstrasse 1, D-85764 Neuherberg, Germany
    Search for more papers by this author
  • NINA BUCHMANN

    1. Institute of Plant, Animal and Agroecosystem Sciences, ETH Zurich, Universitaetsstrasse 2, CH-8092 Zurich, Switzerland
    Search for more papers by this author

M. Gubsch. e-mail: mgubsch@gmail.com

ABSTRACT

Plant and soil nitrogen isotope ratios (δ15N) were studied in experimental grassland plots of varying species richness. We hypothesized that partitioning of different sources of soil nitrogen among four plant functional groups (legumes, grasses, small herbs, tall herbs) should increase with diversity. Four years after sowing, all soils were depleted in 15N in the top 5 cm whereas in non-legume plots soils were enriched in 15N at 5–25 cm depth. Decreasing foliar δ15N and Δδ15N (= foliar δ15N − soil δ15N) values in legumes indicated increasing symbiotic N2 fixation with increasing diversity. In grasses, foliar Δδ15N also decreased with increasing diversity suggesting enhanced uptake of N depleted in 15N. Foliar Δδ15N values of small and tall herbs were unaffected by diversity. Foliar Δδ15N values of grasses were also reduced in plots containing legumes, indicating direct use of legume-derived N depleted in 15N. Increased foliar N concentrations of tall and small herbs in plots containing legumes without reduced foliar δ15N indicated that these species obtained additional mineral soil N that was not consumed by legumes. These functional group and species specific shifts in the uptake of different N sources with increasing diversity indicate complementary resource use in diverse communities.

INTRODUCTION

During the last decade an increasing number of studies in experimental grasslands have observed positive effects of biodiversity on primary productivity (Balvanera et al. 2006; Cardinale et al. 2007). The main mechanisms discussed as explanations for positive biodiversity–productivity relationships are: (1) ‘complementarity effects’ where resource partitioning and facilitation among species result in a more complete use of available resources in plant communities of increasing diversity; and (2) ‘selection effects’ where particular species have disproportionately large effects on the resource use at community level (Tilman et al. 1997; Loreau 1998). However, empirical evidence for resource complementarity and niche separation among plant species is still ambiguous.

Nitrogen (N) is likely to be the most critical nutrient limiting net primary productivity in temperate ecosystems (Vitousek & Howarth 1991). An increasing belowground resource use reflected in reduced soil-extractable nitrate at increasing plant diversity has been reported in several studies with temperate grassland species (e.g. Tilman, Wedin & Knops 1996; Niklaus et al. 2001; Palmborg et al. 2005; Oelmann et al. 2007). Niche separation due to complementary use of nitrogen might be one possible mechanism allowing for the coexistence of many plant species at small scale. Although there is evidence that plant species differ in their capacity to assimilate inorganic and organic N (Miller & Bowman 2003; Weigelt, Bol & Bardgett 2005), less is known how interactions with neighbouring plants affect the use of different N sources, that is, the use of different N forms or the uptake from different soil depths, at increasing plant diversity. Results from 15N-labeling with different N forms suggested that N allocation strategies play an important role (Kahmen et al. 2006) or that N partitioning may allow for niche separation among subordinate plant species (McKane et al. 2002; Aanderud & Bledsoe 2009; von Felten et al. 2009). Moreover, recent studies indicated neighbour-identity effects on plant species capacity to take up different N forms or enhance the uptake of particular N forms (Miller, Bowman & Suding 2007; Ashton et al. 2008).

At natural abundance, stable 15N isotopes in plants and soils have been suggested as a promising and non-destructive tool to assess ecosystem N dynamics and to detect plant species-specific changes in N acquisition and utilization (Handley & Raven 1992; Högberg 1997; Robinson 2001). Plants have access to various N forms in the soil, that is, ammonium, nitrate and organic N, which may differ in their isotopic compositions (Robinson 2001). Isotope effects during ammonium and nitrate assimilation are likely to differ because of differences in the location of ammonium (only roots) and nitrate (roots, shoots) assimilation and the involvement of different assimilatory enzyme systems (glutamine synthetase, nitrate reductase) (Robinson, Handley & Scrimgeour 1998; Kolb & Evans 2003). Although physiological factors during N uptake and assimilation or plant-internal N cycling may modulate plant nitrogen isotopic compositions (Evans 2001), plant nitrogen isotope ratios (δ15N) are still influenced by the isotopic composition of the N source. Discrimination against 15N during different N cycle processes such as N mineralization, nitrification, denitrification or ammonia volatilization lead to soil N pools with different δ15N signatures (Mariotti et al. 1981; Högberg 1997; Robinson 2001). At high N availability, N lost through denitrification and leaching of inorganic N forms is depleted in 15N. Thus, the remaining N pool is ongoingly enriched in 15N, which may be reflected in high foliar δ15N values. In addition, at low N availability, plants are likely to be more dependent on mycorrhizal fungi for N acquisition, where the fractionation during N transfer may result in lower foliar δ15N values (Evans 2001). Nitrogen may also be directly transferred from N2-fixing legumes to neighbouring plants (Haystead, Malajczuk & Grove 1988; Mäder et al. 2000; Govindarajulu et al. 2005). In experimental grasslands, lower foliar δ15N values of non-legumes in mixtures with legumes have been attributed to a transfer of 15N-depleted symbiotically fixed nitrogen (Spehn et al. 2002; Temperton et al. 2007). However, these studies did not account for different 15N signatures in the soil. Higher N availability at high soil N concentrations or due to N2-fixing legume neighbours will affect the relationship between foliar δ15N values and leaf nitrogen concentrations of non-legumes. Thus, the combined analysis of δ15N values in soils and leaves together with N concentrations in leaves could serve as an integrative measure of plant N acquisition from different N sources.

We used a large biodiversity experiment (Jena Experiment, Roscher et al. 2004) to investigate how plant community diversity affected: (1) changes in bulk soil δ15N values four years after establishment compared with initial values; (2) foliar δ15N values and leaf nitrogen concentrations of 51 different plant species assigned to four plant functional groups (legumes, grasses, non-legume tall herbs and small herbs); and (3) foliar δ15N values and leaf nitrogen concentrations of the grass species Lolium perenne L. sown as an additional species in subplots of each experimental mixture. Our major hypothesis was that nitrogen partitioning among the four functional groups should increase with increasing diversity, thus allowing for complementary use of different N sources.

MATERIALS AND METHODS

Study site and experimental design

The field site of the Jena Experiment is located in the floodplain of the river Saale north to the city of Jena (Thuringia, Germany, 50°55′ N, 11°35′ E, 130 m a.s.l.). The Jena Experiment was established in May 2002 on a former arable land, which had been regularly ploughed and fertilized over the last 40 years. The area around Jena has a mean annual temperature of 9.3 °C and a mean annual precipitation of 587 mm (Kluge & Müller-Westermeier 2000). The soil of the experimental site is a Eutric Fluvisol. Due to flooding dynamics the soil texture ranges from sandy loam close to the river to silty clay with increasing distance from the river Saale. Plant communities were established from a pool of 60 species that are common in semi-natural Central European mesophilic grasslands (Molinio-Arrhenatheretea, Ellenberg 1988). Based on differences in their morphology, physiology and phenology, species were classified into four functional groups (12 legumes, 16 grasses, 12 small herbs, 20 tall herbs) using a cluster analysis with trait data derived from the literature (Roscher et al. 2004). The experiment comprises 82 plots of 20 × 20 m size. The design is near-orthogonal whereby the factors species richness (1, 2, 4, 8, 16 and 60) and functional group number (1, 2, 3 and 4) vary as independently as possible with the restriction that plant functional group number cannot exceed species richness in a given mixture. Species composition for replicates of each Species richness × Functional group number combination was determined by independent random draws with replacement from species of the respective functional groups. Total sowing density amounted to 1000 viable seeds per square metre divided in equal proportions among species in mixtures. Plots were assigned to four experimental blocks parallel to the river to account for differences in soil texture. Each of them contained the same number of plots per species-richness level.

In an additional sub-experiment L. perenne (variety ‘Juwel’) was sown on a sub-area of 0.8 × 1.0 m size in each large 20 × 20 m plot (Roscher et al. 2008a). A low number of 100 viable seeds per square metre was added to the 1000 seeds sown for the regular mixture of the plant community. Lolium perenne is a high-yielding grass of good fodder quality, often used in agricultural grassland seed mixtures. It belongs to the functional group of grasses as derived from the literature-based trait cluster analysis (see earlier discussion). Its presence in all plots allowed us to have at least one common ‘phytometer’ across the entire experimental design.

All plots were weeded twice per growing season when the vegetation was low and the canopy not completely closed (early April at the beginning of the growing season and early July after mowing). Species not sown into a particular plot were removed to maintain the sown species combinations. They were mown twice a year (early June and September) according to the traditional management regime of hay meadows in the study region. No fertilizer was applied during the experiment.

Plant sampling and preparation

Bulk samples of the uppermost fully expanded leaves of five to seven shoots were collected along transects (excluding the outer margin of the plots). Sampling took place at estimated peak standing biomass before the first mowing in May 2006 and in all plots (except for one 60 species plot and four monocultures) where species were originally sown and established successfully. Nine out of 60 species were excluded from the study because they had already finished their life cycle in early spring, were sown in few mixtures or were not well established in mixtures. A full list of all species sampled is given in the Supporting Information (Supporting Information Table S1). Additionally, 10–15 fully expanded leaves from at least three different individuals of L. perenne were sampled in all experimental plots where the species was available (see earlier discussion). Leaf samples were dried for 48 h at 70 °C and ground with a ball mill. Leaf nitrogen isotope ratios (i.e. foliar δ15N) and N concentrations were determined from approximately 3 mg with an isotope-ratio mass spectrometer (IRMS; Deltaplus XP and Delta C prototype Finnigan MAT, respectively, Finnigan MAT, Bremen, Germany; 0.1‰ precision). The δ15N values represent nitrogen isotopic composition of the sample relative to that of atmospheric dinitrogen in ‰:

image((1),)

where Rsample is the sample isotope ratio (15N/14N) and Rstandard is the 15N/14N ratio for atmospheric N2.

Soil sampling and preparation

Soil sampling was performed in April 2002 before sowing and repeated in April 2006. In each plot, five soil cores were taken at random locations with a split-tube sampler (inner Ø 4.8 cm; Eijkelkamp Agrisearch Equipment, Giesbeek, Netherlands) in 2002. In a paired sampling design, to avoid additional variation through spatial variability, three of the five sampling locations were chosen for the sampling in 2006. Sampling was carried out at a spatial distance of less than 30 cm to the locations of the first sampling in 2002. Soil samples were taken to a depth of 30 cm, which encompasses the main rooting zone of the experimental species at the field site (Bessler et al. 2009). Soil cores were segmented at a resolution of 5 cm depth, resulting in six subsamples per core. Soil samples were dried at 40 °C. After sieving to 2 mm mesh size, plant debris was removed with tweezers in 2002. Due to the higher proportion of roots in the soil, samples in 2006 were sieved to 1 mm mesh size for standardization (for details see Steinbeiss et al. 2008). Soil nitrogen isotope ratios (i.e. bulk soil δ15N) were measured from 50 mg of dried soil (after grinding with a ball-mill) with an IRMS (Delta C prototype IRMS, Finnigan MAT) and calculated as described earlier (Eqn 1).

Data analyses

Changes in soil δ15N values through time were calculated per plot and per 5 cm segment of the soil profile as differences between soil δ15N values in 2006 and 2002. Variation in foliar Δδ15N values indicates depletion or enrichment in 15N compared with the soil δ15N background and reflects plant-specific patterns of N uptake and use. Thus, foliar Δδ15N values were derived plot-wise for each species as the difference between foliar δ15N and soil δ15N in 2006 to correct for plot-specific differences in background bulk soil δ15N values (Amundson et al. 2003). For these calculations, bulk soil δ15N values were averaged across the profile from 5 to 30 cm depth, excluding the uppermost 5 cm because preliminary analyses (see Results) suggested that this layer strongly reflected the isotopic signature of litter input. However, all analyses were also done with Δδ15N based on soil δ15N values from 0 to 5 cm depth. When deviating results were obtained, they are mentioned in the Results section.

We used mixed-effects models for all analyses based on species-specific data (foliar δ15N, foliar Δδ15N, leaf nitrogen concentrations), applying the lme function implemented in the nlme package (Pinheiro & Bates 2000) of the statistical software R2.6.2 (R Development Core Team, http://www.r-project.org). Data were analysed for each functional group separately to assess plant diversity effects on different species within functional groups. In addition, all non-legume species were combined in additional analyses to test for differences between these as a group. Block and plot identity were treated as random factors in a nested sequence. Starting from a constant null model with the random factors only, fixed effects were entered step-wise in the following order: species richness (SR; log-linear), functional group number (FG; linear), species or functional group identity (ID), respectively, and interaction terms between identity and the experimental factors (ID × SR, ID × FG). We applied the maximum likelihood method and used likelihood ratio tests to assess the statistical significance of model improvement by adding the fixed effects. In alternative models, we replaced the term for functional group number by the presence/absence of particular plant functional groups, although interactions between species identity and the presence/absence of particular plant functional groups could not be tested because of the low number of replicates per combination. Furthermore, we classified non-legumes into species with different root types (prevalence of primary or secondary root system) and rooting depth (see Supporting Information Table S1; Roscher et al. 2004) and tested whether these root morphological characteristics explained a significant amount of variation in foliar δ15N and Δδ15N. Finally, we explored relationships between foliar Δδ15N and leaf nitrogen concentrations for each plant functional group and tested for species-specific differences within plant functional groups.

Analyses of variance (anova) with sequential sums of squares were used for analysing data of L. perenne sown on all experimental plots. Statistical models were structured as described earlier, entering block before the experimental factors. Split-plot anova was applied for analyses of soil δ15N to test for variation dependent on soil depth and to examine whether effects of the experimental factors changed with soil depth.

RESULTS

Soil 15N characteristics

Initial bulk soil δ15N values before the establishment of the biodiversity experiment in 2002 were on average 6.59‰ (± 0.69 SD) and varied largely among the experimental blocks (F3,76 = 42.99, P ≤ 0.001), being more depleted in 15N near to the river. Bulk soil δ15N values in 2002 increased with soil depth (F1,407 = 119.67, P ≤ 0.001), with δ15N values in 0–20 cm soil depth around 6.52‰ (± 0.65 SD) and 6.75‰ (± 0.71 SD) in 20–30 cm soil depth. Four years later, in April 2006, bulk soil δ15N values decreased with increasing number of sown species. Effects of the number of plant functional groups and the presence/absence of particular plant functional groups depended on soil depth (Table 1). Bulk soil δ15N at the soil surface (upper 5 cm) was significantly depleted in 15N compared with deeper soil layers (Fig. 1). The depletion in 15N at the soil surface was significantly larger in plots with legumes than in plots without legumes, but at deeper soil layers this difference disappeared (Fig. 1a). In contrast, bulk δ15N at the soil surface did not vary between plots with and without grasses, but soil was more enriched in 15N at deeper layers when grasses were absent (Fig. 1b). Differences in bulk soil δ15N between 2006 and 2002 showed a significant depletion in 15N in the top 5 cm (Table 1, Fig. 1c & d), with significantly higher depletion of 15N in plots with legumes. In contrast, soil in non-legume plots became significantly enriched in 15N at soil depth from 5 to 25 cm in 2006 compared with initial values in 2002 (Fig. 1c). This effect did not depend on the presence of a particular non-leguminous plant functional group. Species and plant functional group richness did not affect changes in soil δ15N from 2002 to 2006 (P > 0.050).

Table 1.  Summary of analysis of variance (anova) of bulk soil δ15N values measured at six different soil depths (from 0 to 30 cm) in samples taken four years after establishment of the biodiversity experiment (April 2006), and changes in bulk soil δ15N values between 2006 and the year of sowing (April 2002)
Source of variationd.f.Soil δ15N (2006)Change soil δ15N (2006–2002)
MSFPMSFP
  1. Given are the degrees of freedom (d.f.), mean sums of squares (MS), F ratios (F) and P values (P). Note that contrasts for the presence/absence of particular plant functional groups were fitted in series of analyses replacing the functional group term. Experimental factors were tested against plot identity. Soil depth and its interactions with the experimental factors were tested against the residuals. Significant effects are marked in bold. Arrows indicate a significant decrease (↓) or increase (↑) of analysed variables with increasing species richness, presence of particular plant functional groups or soil depth.

Block331.56343.28<0.0011.1453.450.021
Species richness (SR; log-linear)14.1085.630.0200.7082.130.148
Functional group number (FG; linear)10.2040.280.5990.7872.370.128
Legume presence (LE)10.001<0.010.9692.9549.730.003
Grass presence (GR)10.4960.680.4110.0430.120.725
Small herb presence (SH)10.8171.130.2910.0020.010.943
Tall herb presence (TH)10.6200.860.3580.1010.300.588
Plot760.792  0.332  
Depth (linear)111.758301.08<0.0011.75724.33<0.001
Depth × SR10.0401.020.3130.0010.010.915
Depth × FG10.1684.300.0390.2313.200.074
Depth × LE11.07929.32<0.0010.7029.870.002
Depth × GR10.3071.940.0050.1381.910.168
Depth × SH10.0020.050.8200.1311.810.179
Depth × TH10.0411.030.311<0.001<0.010.971
Residuals4070.039  0.072  
Figure 1.

Depth profile of bulk soil δ15N values 4 years after sowing in April 2006 (a) for plots with/without legumes, and (b) for plots with/without grasses as well as changes of bulk soil δ15N values in comparison with the year of sowing (2006–2002) (c) for plots with/without legumes, and (d) with/without grasses. Values are means across plots ± 1 SE. Closed circles represent plots with legumes (a, c) or grasses (b, d), respectively. Open circles represent plots without species of the respective plant functional groups.

Foliar 15N characteristics

Foliar δ15N values in May 2006 differed significantly among non-legume plant functional groups (Table 2), reaching average values of 2.21‰ (± 0.85 SD) in grasses, 1.82‰ (±1.35 SD) in tall herbs and 1.36‰ (±1.10 SD) in small herbs. Average foliar δ15N of legumes were −0.54‰ (± 0.65 SD). Foliar Δδ15N values were negative for all plant species, showing pronounced depletion in 15N compared with bulk soil N. N isotopic signatures of legume species varied independently from soil δ15N (L = 1.06, P = 0.303; Fig. 2a), whereas foliar δ15N values increased with soil δ15N values for grasses, tall herbs and small herbs (L = 15.20, P ≤ 0.001). Although the absolute differences between foliar and soil δ15N values varied among non-legume plant functional groups (L = 37.99, P ≤ 0.001) and among species (L = 187.74, P ≤ 0.001), the slope of the relationships between foliar and soil δ15N values differed neither among non-legume plant functional groups (L = 2.82, P = 0.244) nor species (L = 36.31, P = 0.593; Fig. 2b–d).

Table 2.  Summary of mixed-effects model analyses of foliar δ15N and Δδ15N values (i.e. foliar δ15N – soil δ15N) for all non-legumes in May 2006
 Non-legumes 
Foliar δ15NFoliar Δδ15
LPLP
  1. Differences among species grouped into different non-legume plant functional groups and classified according to their root type or rooting depth (Supporting Information Table S1) were tested in separate model series after fitting the main effects. Models were fitted by stepwise inclusion of fixed effects. Likelihood-ratio tests were applied to assess model improvement (L ratio) and the statistical significance of the explanatory terms (P values). Significant effects are marked in bold. Arrows indicate a significant decrease (↓) of the measured variables with increasing species richness, functional group number or the presence of particular plant functional groups. Note that contrasts for the presence/absence of particular plant functional groups were fitted in series of analyses replacing the functional group term.

Species richness (SR; log-linear)4.400.0360.510.475
Functional group number (FG; linear)1.560.2121.010.316
Legume presence (LE)1.240.2662.920.087
Functional group identity    
 Functional group identity (ID)34.82<0.00138.48<0.001
 Functional group ID × SR6.800.03310.780.005
 Functional group ID × FG1.670.4351.190.553
Root depth    
 Root depth15.930.00314.900.005
 Root depth × SR7.510.1117.540.110
 Root depth × FG10.780.02913.110.011
Root type    
 Root type2.920.2334.040.133
 Root type × SR9.020.01110.030.007
 Root type × FG0.290.8640.410.815
Figure 2.

Foliar δ15N values in May 2006 as a function of bulk soil δ15N values for (a) legumes (b) grasses (c) small herbs, and (d) tall herbs as well as a function of sown species richness for (e) legumes (f) grasses (g) small herbs and (h) tall herbs, and foliar Δδ15N values (i.e. foliar δ15N–soil δ15N) as a function of sown species richness for (i) legumes (k) grasses (l) small herbs, and (m) tall herbs. Significant relationships are indicated with continuous lines (P < 0.050), non-significant relationships are shown with broken lines.

Foliar δ15N and Δδ15N values of legumes and non-legumes decreased with increasing species richness (Tables 2 & 3, Fig. 2e–h). Functional group number and tall herb presence also affected foliar δ15N of legumes, but their effects did not remain statistically significant after correcting for background soil δ15N (Δδ15N; Table 3). The overall significant species-richness effect on foliar δ15N values of non-legumes became also insignificant after correcting for soil δ15N (Δδ15N; Table 2). However, the effects of increasing species richness on both, foliar δ15N and Δδ15N values, depended on the functional group identity of non-legumes (Table 2).

Table 3.  Summary of mixed-effects model analyses of foliar δ15N values, Δδ15N values (foliar δ15N – soil δ15N) and leaf nitrogen concentrations for each plant functional group in May 2006
Source of variationFoliar δ15NFoliar Δδ15NLeaf N concentrationFoliar δ15NFoliar Δδ15NLeaf N concentration
LPLPLPLPLPLP
  1. Models were fitted by stepwise inclusion of fixed effects. Likelihood-ratio tests were applied to assess model improvement (L ratio) and the statistical significance of the explanatory terms (P values). Significant effects are marked in bold. Arrows indicate a significant increase (↑) or decrease (↓) of the measured variables with increasing species richness, functional group number or the presence of particular plant functional groups. Note that contrasts for the presence/absence of particular plant functional groups were fitted in series of analyses replacing the functional group term.

 LegumesGrasses
Species richness (SR; log-linear)12.67<0.0014.640.0312.600.10710.680.0016.530.0110.880.348
Functional group number (FG; linear)5.890.0152.290.1300.090.7701.150.2842.690.1010.060.807
Legume presence (LE)2.420.1204.240.0401.460.227
Grass presence (GR)0.710.4010.010.9390.640.425
Small herb presence (SH)2.500.1142.190.1391.870.1710.450.5010.540.4610.290.588
Tall herb presence (TH)5.500.0192.380.123<0.010.997<0.010.9710.230.6331.610.205
Species identity (ID)76.14<0.00179.33<0.001114.46<0.00146.64<0.00148.38<0.00119.820.071
Species ID × SR19.360.03613.800.18225.300.00519.070.08716.590.16612.180.431
Species ID × FG7.720.6569.170.5165.530.85339.92<0.00135.81<0.00130.300.003
 Small herbsTall herbs
Species richness (log-linear)4.100.0430.820.3660.360.5460.670.4141.090.2971.120.291
Functional group number (linear)0.170.683<0.010.971<0.010.9820.520.4690.720.3950.020.886
Legume presence (LE)0.010.9240.050.8276.090.0140.050.8190.190.6628.320.004
Grass presence (GR)0.430.5101.250.2633.350.067<0.010.9700.040.8514.280.039
Small herb presence (SH)1.700.1921.500.2210.130.713
Tall herb presence (TH)1.940.1631.590.2070.120.729
Species identity (ID)35.46<0.00135.81<0.00141.19<0.00150.53<0.00152.64<0.00173.60<0.001
Species ID × SR8.200.6958.690.65115.120.17731.010.00629.670.00912.370.577
Species ID × FG41.20<0.00148.51<0.00116.750.11617.510.23011.450.65039.34<0.001

Species root depth or root type (Supporting Information Table S1) entered as covariate in statistical analyses did not explain differences in foliar δ15N and Δδ15N values of non-legume species assigned to different plant functional groups (analyses not shown). However, foliar δ15N and Δδ15N values of non-legume species and their variation in response to increasing functional group number varied among species with different root depths, where species with deeper roots were enriched in 15N. Variation in foliar δ15N and Δδ15N values of non-legumes was not dependent on species root type, but their response to increasing species richness differed among non-legume species with different root types (Table 2).

Analyses at the functional group level showed that grasses decreased their foliar δ15N and Δδ15N values with increasing plant species richness, and that grasses were more depleted in 15N when growing in plant communities with legumes compared with non-legume communities (Table 3, Fig. 2f, k). The legume effect did not remain significant when top 5 cm soil δ15N were used to calculate foliar Δδ15N. Small herbs also decreased foliar δ15N values with increasing species richness. However, Δδ15N values of small and tall herbs were not influenced by species richness, functional group number or the presence of particular plant functional groups (Table 3, Fig. 2g–h, l–m). Even though we found functional group-specific patterns, foliar δ15N and Δδ15N values varied significantly among species within plant functional groups (Table 3). Effects of increasing species richness (in tall herbs) and functional group number (in grasses and small herbs) on foliar δ15N and Δδ15N values were also species-specific (significant interactions ID × SR, ID × FG, Table 3). To visualize species differences in foliar Δδ15N values and their variation in response to increasing species richness, we plotted species mean Δδ15N values and slopes of regressions against species richness in Fig. 3.

Figure 3.

Slopes of regressions of foliar Δδ15N values (i.e. foliar δ15N–soil δ15N) against sown species richness (±1 SE) plotted against mean values (±1 SE) of foliar Δδ15N values for individual species (a) legumes (b) grasses (c) small herbs, and (d) tall herbs sampled in May 2006. Cases above the broken line indicate that foliar Δδ15N values increased with increasing species richness (positive slopes), whereas cases below the broken line indicate that foliar Δδ15N values decreased with increasing species richness (negative slopes). Note that one tall herb species (Centaurea jacea) with Δδ15N values (−3.40‰ ± 1.18) and an extraordinary large positive regression slope against species richness (4.52 ± 2.75) was not plotted.

Leaf nitrogen concentrations of legumes and their response to increasing species richness varied among species (Table 3) and were higher at larger foliar Δδ15N values (L = 9.38, P = 0.002; Fig. 4a). Leaf nitrogen concentrations of grasses (L = 3.26, P = 0.071; Fig. 4b) and small herbs (L = 2.51, P = 0.113; Fig. 4c) varied independently from foliar Δδ15N values. Leaf nitrogen concentrations of tall herbs were positively related to foliar Δδ15N values (L = 16.25, P ≤ 0.001), but this relationship depended strongly on species identity (L = 35.48, P ≤ 0.001; Fig. 4d). Leaf nitrogen concentrations varied significantly among different small herb and tall species (Table 3). Variation of leaf nitrogen concentrations of tall herb species in response to increasing functional group number depended on tall herb species identity. However, in general small and tall herb leaf nitrogen concentrations were higher in the presence of legumes, whereas leaf nitrogen concentrations of tall herbs were lower in the presence of grasses (Table 3). Leaf nitrogen concentrations of grasses did not differ significantly among species, but their variation in response to increasing functional group number was species dependent (Table 3).

Figure 4.

Leaf nitrogen concentrations (±1 SE) plotted against foliar Δδ15N values (i.e. foliar δ15N–soil δ15N) for individual species of (a) legumes, (b) grasses, (c) small herbs, and (d) tall herbs. Values are means (±1 SE) across all plant communities where a particular species was present in May 2006. Note different scaling at the y-axis for legumes.

Foliar 15N characteristics of Lolium perenne

Foliar δ15N of L. perenne, grown as an additional species in all experimental plots, correlated positively with soil δ15N (R = 0.396, P ≤ 0.001, n = 72; Fig. 5a), and similarly to Δδ15N, became depleted in 15N with increasing species richness and functional group number (Table 4, Fig. 5b & c). Although Lolium perenne had significantly lower foliar δ15N values in the presence of other grasses and tall herbs, the effects of other grasses were not statistically significant after correcting for soil δ15N (Δδ15N; Table 4). When foliar δ15N values were related to soil δ15N at 0–5 cm depth, the presence of legumes resulted in a decrease in Δδ15N of L. perenne. Leaf nitrogen concentrations of L. perenne did not correlate with Δδ15N values (R = 0.222, P = 0.061, n = 72), whereas legume presence had positive effects and the presence of other grasses had negative effects on leaf nitrogen concentrations of L.perenne (Table 4).

Figure 5.

Foliar δ15N values as a function of bulk soil δ15N values (a), foliar δ15N values as a function of sown species richness (b), and Δδ15N values (i.e. foliar δ15N – soil δ15N) in May 2006 as a function of sown species richness (c) for Lolium perenne sown as an additional species in all experimental plots.

Table 4.  Summary of analysis of variance (anova) of foliar δ15N values, Δδ15N values (foliar δ15N – soil δ15N) and leaf nitrogen concentrations for Lolium perenne sown as an additional species in all experimental plots for May 2006
Source of variationd.f.Foliar δ15NFoliar Δδ15NLeaf N concentration
MSFPMSFPMSFP
  1. Given are the degrees of freedom (d.f.), mean sums of squares (MS), F ratios (F) and P values (P). Note that contrasts for the presence/absence of particular plant functional groups were fitted in series of analyses replacing the functional group term. Significant effects are marked in bold. Arrows indicate a significant decrease (↓) or increase (↑) of analysed variables with increasing species richness or presence of particular plant functional groups.

Block32.2904.190.0090.9141.530.21618.170.470.707
Species richness (log-linear, SR)114.03626.61<0.0018.94014.91<0.0019.320.240.627
Functional group number (linear, FG)13.3206.510.0162.6914.490.0380.07<0.010.967
Legume presence (LE)10.3570.600.4410.1410.220.640989.9541.31<0.001
Grass presence (GR)12.8695.170.0261.6982.760.101419.0312.850.001
Small herb presence (SH)10.9241.580.2130.2890.450.50212.800.330.576
Tall herb presence (TH)12.3464.170.0453.0895.210.02675.181.990.163
Residuals660.548  0.599  38.96  

DISCUSSION

Soil δ15N values

Bulk soil δ15N values at the experimental site ranged between 5 and 9‰, which is quite common for agricultural sites but were slightly more enriched than known from other grassland studies (Watzka, Buchgraber & Wanek 2006; Kahmen, Wanek & Buchmann 2008). Closer to the river these values became more depleted (Table 1). The experimental site has been used as arable land 40 years prior to the Jena Experiment and is embedded in regularly fertilized agricultural land. Leaching of mostly 15N-depleted inorganic fertilizer (Choi, Ro & Hobbie 2003; Watzka et al. 2006) from agricultural land and differences in soil texture across the experimental site with higher clay contents (absorbing 15N-enriched NH4+) with increasing distance from the river may explain this spatial variation perpendicular to the river.

Generally, changes in bulk soil δ15N values during the 4-year study period from 2002 to 2006 were rather small (0.27 ± 0.01‰ averaged across all plots and soil layers). The observed small changes are attributable to the large total soil N stocks (846.5 ± 10.2 g N m−2 in 2006 for 0–30 cm soil depth and averaged across all plots), which hamper that biological processes discriminating against 15N have large effects on bulk soil δ15N values. Increasing species richness correlated with a depletion in 15N of bulk soil sampled four years after sowing but species richness did not significantly affect changes in bulk soil δ15N values compared with the initial soil δ15N values in 2002 (Table 1). Nevertheless, changes in bulk soil δ15N values in the soil profile strongly differed between plots with legumes and non-legume plots (Fig. 1c). The top soil (upper 5 cm) was generally depleted in 15N, with considerably lower δ15N values in plots with legumes due to the input of 15N-depleted legume litter. Application of the 15N natural abundance method has shown that legume species on average derived more than 70% of their nitrogen from symbiotic N2 fixation in our experiment (Roscher et al. 2010). The role of litter input for a depletion in 15N in the upper 5 cm of the soil profile is supported by the fact that soil N was accumulated in the upper 15 cm of the soil profile from 2002 to 2006, whereas a loss of soil N was observed in the deeper layers of the soil N profile (20–30 cm) (unpublished data). Over time, deeper soil layers (5–25 cm) became enriched in 15N in non-legume plots, whereas this enrichment was negligible in plots with legumes. Biogeochemical processes that determine soil δ15N signatures are spatially and temporally highly variable (Högberg 1997; Robinson 2001) and multiple processes, for example, ammonification, nitrification, N uptake by plants and microorganisms, denitrification, leaching and volatilization, interact simultaneously (Schimel & Bennett 2004). A 15N enrichment of bulk soil has mostly been attributed to higher rates of soil N cycling associated with losses of mineral N containing the lighter isotopic forms, leaving the remaining soil enriched in 15N (Mariotti et al. 1981; Pardo et al. 2006). Inorganic nitrogen may be absorbed to clay minerals (NH4+), taken up by plants or microbes or leached (NO3-) into the groundwater. Higher litter decay, nitrification rates and nitrate leaching have been shown to be associated with high legume abundances (Scherer-Lorenzen et al. 2003). Soil-extractable nitrate concentrations were higher in plant communities with legumes and reduced to lower levels in plant communities with grasses in the Jena Experiment (Oelmann et al. 2007; Roscher et al. 2008b). In contrast, soil-extractable ammonium concentrations were often below the detection limit at our study site (Oelmann et al. 2007). In this context, smaller changes in bulk soil δ15N values in 5–25 cm depth in plots with legumes suggest that the overall loss of 15N-depleted nitrogen and the subsequent enrichment of bulk soil δ15N were compensated by the input of 15N-depleted legume litter throughout the top 25 cm soil profile. In contrast, increasing soil δ15N values in non-legume communities reflected the expected 15N enrichment of soil organic matter due to nitrification and N uptake by plants and microorganisms. The observed variation in soil δ15N strongly supports the need to analyse plant δ15N relative to soil δ15N, that is, Δδ15N, when comparing across different sites.

Functional group and species differences in foliar δ15N and Δδ15N values

Foliar δ15N values are affected by a range of factors that might differ in their magnitude dependent on plants' internal and external conditions: (1) multiple N sources and N forms with distinct 15N compositions (dry and wet deposition, N2 fixation, organic or inorganic soil N); (2) different soil depths from which plants take up N; (3) mycorrhizal associations; (4) fractionation processes during symbiotic N2 fixation of legumes or N uptake and assimilation; and (5) plant internal N allocation and partitioning during metabolic transformations (Shearer & Kohl 1986; Yoneyama et al. 1991; Handley & Raven 1992; Högberg 1997; Evans 2001). The various biological processes that fractionate to some degree against 15N may result in a deviation of plants' δ15N signatures from the source. For instance, discrimination rates of 15 and 17‰ have been reported for enzymes involved in the assimilation of ammonium and nitrate (nitrate reductase, glutamine synthetase), whereas symbiotic N2 fixation via nitrogenase leads to N isotope fractionation of 0–6‰ (Robinson 2001). Thus, the interpretation of experimental field data remains challenging because several processes simultaneously affect bulk δ15N values of plant tissue.

Foliar δ15N values of grasses, small and tall herbs correlated positively with bulk soil δ15N values. Regression slopes between foliar and soil δ15N values that did not differ significantly among functional groups indicated no general differences in the reliance on soil nitrogen of the tested plants. Furthermore, foliar Δδ15N values for all species were negative across all plant diversity levels. Negative Δδ15N values of non-leguminous grassland species have also been reported by Kahmen et al. (2008) and are often attributed to mineral soil uptake, which is depleted in 15N compared with bulk soil N (Robinson 2001). However, discrimination during ammonium and nitrate absorption and assimilation could also affect the isotopic signature of assimilated N relative to the source, but this is more likely to occur at high concentrations of NH4+ and NO3- in the soil solution (Kolb & Evans 2003).

Our analyses at the functional group level (Table 2) as well as separate analyses of non-legume plant functional groups (Table 3) showed that average foliar Δδ15N values differed among plant functional groups and among species within functional groups (Fig. 4). Foliar Δδ15N values, which correct for differences in soil δ15N background, reflect species-specific patterns in the uptake of different mineral N sources. Grasses were generally more enriched in 15N (highest foliar δ15N and Δδ15N) compared with herbs. In semi-natural temperate grasslands, Kahmen et al. (2008) have shown that foliar Δδ15N values of 15 grassland species correlated negatively with decreasing NH4+ uptake and positively with increasing ratios of NO3- versus NH4+ uptake, which would indicate on average greater reliance of grasses on nitrate as mineral N source. A greater nitrate assimilation of grasses could also increase their foliar Δδ15N values, when partial reduction of NO3- in the roots by nitrate reductase results in a 15N-enriched NO3- pool, which is transported to the shoot for further assimilation (Evans et al. 1996; Kolb & Evans 2003). Mycorrhizal fungi, which preferentially transfer isotopically depleted nitrogen to their host plants, may also affect plants δ15N signatures (Högberg 1997; Hobbie, Macko & Williams 2000; Craine et al. 2009). Grassland species are usually infected with arbuscular mycorrhizal fungi (AMF), although plants with more fine roots tend to be less infested with mycorrhizal fungi (Brundrett 2002). Thus, a less abundant mycorrhizal association could also be responsible for the higher δ15N value of grasses with their extensive secondary root system. Among the investigated species only Rumex acetosa L. (Polygonaceae) might be non-mycorrhizal (Brundrett 2009). Indeed, this tall herb had extraordinary large foliar δ15N values in our experiment. Therefore, we cannot exclude effects of differences in mycorrhizal associations on foliar Δδ15N values for different plant species although discrimination of AMF plants against 15N during N uptake is supposed to be marginal (Craine et al. 2009).

Plant diversity effects on foliar δ15N and Δδ15N values

In our study, species-richness effects on non-legume foliar Δδ15N values were largely species-dependent. However, although foliar Δδ15N values of tall and small herbs did not change in response to increasing species richness, grasses including L. perenne generally decreased foliar Δδ15N values (Tables 3 & 4, Figs 3k–l & 5c). These changes indicate shifts in the utilization of different N sources as well as increasing N partitioning among plant species with increasing species richness. Lolium perenne is known to prefer inorganic N compared with organic N, although it is capable to take up organic N compounds (Weigelt et al. 2003, 2005). The uptake of organic N forms by plants is of minor importance at our experimental site as it has been shown that the availability of dissolved organic nitrogen such as free amino acids is rather low (Oelmann et al. 2007; Sauheitl et al. 2010), and microbes are probably the better competitors for organic N sources. Bailey (1998) found that L. perenne absorbed both mineral N forms at equal rates under stress, for example, after defoliation, which was interpreted as a maximization of N acquisition because N uptake was not restricted by a particular N form. In the Jena Experiment, L. perenne becomes light and N limited with increasing species richness (Roscher, Kutsch & Schulze 2011a). Thus, a shift in N partitioning with increasing species and functional group number as well as in the presence of tall herb species would support these findings (Table 4). Generally, foliar δ15N values are assumed to increase at increasing N supply (Garten & van Miegroet 1994; Craine et al. 2009). Kahmen et al. (2008) argued that in managed temperate grasslands with high denitrification rates, NO3- can become enriched in 15N compared with NH4+, which could explain a foliar 15N enrichment with increasing NO3- uptake. Thus, grasses in our study probably increased the uptake of NH4+ relative to NO3- with increasing species richness and lower N availability.

Not surprisingly, foliar δ15N values of legumes did not correlate with bulk soil δ15N values (Fig. 2a). The observed decrease of foliar δ15N and Δδ15N values of legumes with increasing species richness (Fig. 2e, i) was a further indication for increasing N2 fixation of legumes with increasing plant diversity as estimated in a previous study based on the 15N natural abundance method (Roscher et al. 2011b). Legumes are able to take up soil-derived N in addition to the energy-demanding symbiotic fixation of N2, but the utilization of soil nitrogen is more likely when N availability exceeds plant N demands and competition for N is low (Högberg 1997).

Relationships between foliar Δδ15N values and leaf nitrogen concentrations

At global scale foliar δ15N values correlate positively with leaf nitrogen concentrations, whereas several independent studies at regional scale have observed positive relationships between foliar δ15N values and soil N availability (Craine et al. 2009). Generally, a higher N supply is likely to permit greater discrimination against 15N, when the rate of uptake increases assimilation (Evans 2001; Bloom et al. 2010). Our analyses showed highly variable relationships between foliar Δδ15N values (and equally foliar δ15N values) and leaf nitrogen concentrations. A positive relationship between leaf nitrogen concentrations and foliar Δδ15N values was only observed for tall herbs. Although legume presence did not affect foliar δ15N values of herbaceous species, increasing leaf nitrogen concentrations of small and tall herbs in plots with legumes suggested beneficial effects of legumes on their nitrogen supply. Increased soil nitrate concentrations in plots with legumes (Oelmann et al. 2007) stimulate denitrification. Thus, the effects on foliar δ15N value are balanced by the compensation of greater NO3- over NH4+ uptake (resulting in 15N enrichment) with a provision of 15N-depleted legume-derived nitrogen (resulting in 15N depletion). This explains the improved nitrogen nutrition of herbs in plots with legumes without changes in foliar δ15N value. In contrast, grasses showed a significant decrease of foliar Δδ15N values in the presence of legumes (Table 3), that is, larger depletion of foliar δ15N values compared with bulk soil. Grasses therefore predominantly used 15N-depleted legume N directly. The insignificant effects on leaf nitrogen concentrations of grasses do actually not exclude beneficial effects of legumes on N provision to grasses, when additional legume-derived nitrogen is used for growth. Plant nitrogen concentrations may remain constant due to higher biomass production (Roscher et al. 2008b).

CONCLUSIONS

Our study with 51 different grassland species showed that variation in foliar Δδ15N values in response to species richness or functional group number are largely species-specific. Our results clearly demonstrate that the different abilities of plant species to adjust N uptake patterns in response to increasing numbers of neighbouring species contribute to a more complete use of belowground resources at increasing plant diversity. Nevertheless, our study of plant and soil δ15N values may not unambiguously provide a mechanistic explanation for the observed differences in foliar δ15N signatures of the investigated grassland species and the several processes contributing to changes in soil δ15N values over time. Further specific experiments using isotopic tracers are required for a more detailed understanding of plant diversity effects on nitrogen partitioning among plant species and soil N cycling.

ACKNOWLEDGMENTS

The Jena Experiment is funded by the German Science Foundation (FOR456) with additional support from the Max Planck Society and the University of Jena. We acknowledge W. W. Weisser for coordination of the Jena Experiment. We thank all people helping with the management of the experiment, especially A. Weigelt as field coordinator and the gardeners. Many student helpers assisted in soil and plant sampling campaigns and sample preparation. Special thanks to W. Brand, H. Geilmann, K. Sörgel and R. Werner for stable isotope analyses. We thank two anonymous reviewers for their valuable comments that helped to improve the paper.

Ancillary

Advertisement