Evidence from the real world: 15N natural abundances reveal enhanced nitrogen use at high plant diversity in Central European grasslands

Authors


Summary

  1. Complementarity that leads to more efficient resource use is presumed to be a key mechanism explaining positive biodiversity–productivity relationships but has been described solely for experimental set-ups with controlled environmental settings or for very short gradients of abiotic conditions, land-use intensity and biodiversity. Therefore, we analysed plant diversity effects on nitrogen dynamics across a broad range of Central European grasslands.
  2. The 15N natural abundance in soil and plant biomass reflects the net effect of processes affecting ecosystem N dynamics. This includes the mechanism of complementary resource utilization that causes a decrease in the 15N isotopic signal. We measured plant species richness, natural abundance of 15N in soil and plants, above-ground biomass of the community and three single species (an herb, grass and legume) and a variety of additional environmental variables in 150 grassland plots in three regions of Germany. To explore the drivers of the nitrogen dynamics, we performed several analyses of covariance treating the 15N isotopic signals as a function of plant diversity and a large set of covariates.
  3. Increasing plant diversity was consistently linked to decreased δ15N isotopic signals in soil, above-ground community biomass and the three single species. Even after accounting for multiple covariates, plant diversity remained the strongest predictor of δ15N isotopic signals suggesting that higher plant diversity leads to a more closed nitrogen cycle due to more efficient nitrogen use.
  4. Factors linked to increased δ15N values included the amount of nitrogen taken up, soil moisture and land-use intensity (particularly fertilization), all indicators of the openness of the nitrogen cycle due to enhanced N-turnover and subsequent losses. Study region was significantly related to the δ15N isotopic signals indicating that regional peculiarities such as former intensive land use could strongly affect nitrogen dynamics.
  5. Synthesis. Our results provide strong evidence that the mechanism of complementary resource utilization operates in real-world grasslands where multiple external factors affect nitrogen dynamics. Although single species may differ in effect size, actively increasing total plant diversity in grasslands could be an option to more effectively use nitrogen resources and to reduce the negative environmental impacts of nitrogen losses.

Introduction

Concerns about increasing species extinction rates world-wide have set a major focus of current ecological research on the environmental consequences of biodiversity loss (e.g. Scherer-Lorenzen 2005; Naeem, Duffy & Zavaleta 2012). In grassland ecosystems, experimental manipulation of plant diversity revealed significant relationships between plant taxonomic or plant functional group diversity and measures of ecosystem functioning such as productivity (e.g. Hooper et al. 2005; Cardinale et al. 2006; Jax 2010). However, the interpretation of these results is vividly debated, and the underlying mechanisms behind the observed patterns are still poorly understood. Whereas critics attributed the observed effects of biodiversity on ecosystem functioning to hidden treatments and flawed experimental designs (e.g. Huston 1997; Lepš 2004; Thompson & Starzomski 2007), proponents suggested resource facilitation and niche complementarity as driving mechanisms (Tilman, Wedin & Knops 1996; Loreau & Hector 2001; Roscher et al. 2005; Marquard et al. 2009). Even under the low-dimensional conditions of biodiversity experiments where diversity effects are decoupled from major external factors (which could also directly affect resource cycling processes in natural environments), the hypothesis of complementary resource utilization is still discussed (van Ruijven & Berendse 2005; Duffy 2009; Gubsch et al. 2011; Bessler et al. 2012).

In temperate ecosystems, nitrogen (N) is likely to be the most important resource limiting plant growth (Vitousek & Howarth 1991).The way in which plants obtain N strongly affects competitive interactions within plant communities (Wilson & Tilman 1991). The natural abundance of stable 15N isotopes (δ15N) in soils and plants is considered to be a powerful integrator of N cycle processes (for review, see Robinson 2001). Isotopic signals can be used to understand interactions between plants or plant communities and the environment and provide insight into the functional value of biodiversity for ecosystem functioning (Kahmen et al. 2006; Gubsch et al. 2011). However, an important drawback in interpretation of δ15N signals is the fact that multiple interacting factors may be responsible for the isotopic patterns observed. Controls over plant δ15N signals include source differences, several fractionation mechanisms during uptake and specific temporal and spatial N acquisition strategies (Hobbie & Högberg 2012).

In natural and managed grasslands, the type and intensity of land use strongly affect nutrient cycling (Stoate et al. 2001), potentially in ways that differ from those identified in artificial experimental settings (Rose & Leuschner 2012). In this context, an N surplus due to mineral and organic fertilizer application results in higher δ15N values of soil and plants, because mainly the lighter 14N is lost (Watzka, Buchgraber & Wanek 2006). In particular, in the case of slurry or manure application, a significant enrichment of δ15N occurs, because of subsequent ammonia volatilization, which is accompanied by a strong fractionation against the heavier δ15N (Högberg 1997). Beside fertilizer application, grazing and mowing intensity are among to the key parameters that determine land-use intensity in permanent grasslands (Blüthgen et al. 2012). Urine and dung deposition by livestock and the physical disturbance of soil layers by herbivores are supposed to enhance losses of 15N-depleted nitrogen via ammonia volatilization and leaching, leaving the remaining N pool enriched (Aranibar et al. 2008). The removal of biomass by mowing and grazing might also contribute to enhanced losses of the lighter 14N as plant biomass is usually more depleted in δ15N compared to the soil (Kriszan et al. 2009). Thus, high losses of lighter 14N cause overall enrichment of δ15N in ecosystems and provide an indication of the openness of the N cycle, whereas depletion in δ15N suggests a more complete uptake (Ometto et al. 2006; Watzka, Buchgraber & Wanek 2006).

Potential mechanisms identified in biodiversity experiments have to be contrasted with field data derived from observational studies under complex environmental conditions to provide general evidence that biodiversity effectively matters to the functioning of real-world ecosystems (Caliman et al. 2010). Attempts to test the hypothesis of complementary resource utilization in established grasslands, however, have thus far been restricted to very short gradients of abiotic conditions, land-use intensity and biodiversity (e.g. Kahmen et al. 2006; Ashton et al. 2010). Studies that account for multiple drivers of N dynamics across a wide range of environmental condition are lacking.

Here, we studied the plant diversity, natural abundance of 15N in plants and soil, above-ground biomass of the community and three single species (an herb, grass and legume), and a large set of other environmental factors on 150 permanent grassland plots in three regions. Using this comprehensive data set, we wanted to explore the relationship between plant diversity and N uptake across a broad range of environmental conditions, and in particular, to determine whether there is evidence for complementarity of resource uptake among plant species in real-world grasslands.

Materials and methods

Study design

We studied permanent grasslands in three regions in Germany belonging to the Biodiversity Exploratories project (Fischer et al. 2010): the biosphere reserve Schorfheide-Chorin in Brandenburg (NE Germany), the Hainich-Dün area in Thuringia (Central Germany) and the biosphere reserve Schwäbische Alb in Baden-Württemberg (SW Germany). The three study areas reflect a climatic gradient of increasing precipitation, rising altitude and slightly decreasing annual mean temperatures from north to south. While at Schwäbische Alb and Hainich-Dün, all grassland plots are situated on mineral soils, in Schorfheide-Chorin drained fen soils rich in organic matter, such as Gleysols and Histosols, also occur.

In each of the three regions, 48–50 grassland plots of 50 m × 50 m were selected. Type and intensity of land use varied strongly among grasslands, especially regarding mowing regime, amount and type of fertilizer used and grazing management (Blüthgen et al. 2012), but were comparable among regions (Table S1 in Supporting Information). The described land-use gradient is representative of large parts of the Central European grassland vegetation, ranging from unfertilized calcareous dry grasslands to mesotrophic grasslands, and improved fen meadows and pastures, to heavily fertilized silage meadows with several cuts per year (Fischer et al. 2010).

Land-use survey

For each grassland plot, farmers and landowners were interviewed annually between 2006 and 2010 to record the amount (kg N ha−1) and type (mineral, organic or both) of fertilizer applied, and the frequency of mowing and the breeds and density of livestock (livestock units × days ha−1) managed (see Blüthgen et al. 2012 for details). Each variable was then averaged over the entire period.

Fieldwork and chemical analyses

Soil sampling was conducted in early May 2011. On each plot, composite samples of 14 topsoil cores (0–10 cm depth) were collected using a split-tube sampler with a diameter of 5 cm. Cores were taken along two 20-m transects across the respective plot, and roots were removed from the samples in the field. Soil samples were air-dried, sieved to < 2 mm and ground to fine powder. For plant sampling in the 48–50 plots per region, we estimated the cover of all vascular plant species on 4 m × 4 m from mid-May to mid-June 2011 simultaneously in all regions and calculated plant diversity as the number of plant species per 16 m2. Cover sums of the functional groups grasses, legumes and herbs were calculated by summing up the cover of all species in the respective groups. Furthermore, above-ground biomass was harvested in four randomly placed quadrats of 0.25 m2 adjacent to the 4 m × 4 m area used to sample vegetation. Furthermore, we sampled above-ground biomass of at least 20 individuals of each of three plant species belonging to different functional groups (Dactylis glomerata: grass; Trifolium repens: legume; Taraxacum sect. ruderalia: herb) along 50-m transects at each plot. Single plant species biomass was collected whenever present on 50 plots in Hainich-Dün and on nine plots at Schwäbische Alb and Schorfheide-Chorin. Temporary fencing ensured that plots were not mown or grazed before sampling. All plant material was dried for 48 h at 80 °C, weighed and finally ground to fine powder for laboratory analyses.

Nitrogen isotopic composition is referenced to the N standard of atmospheric air (Högberg 1997), which is 15N/14N ≈ 0.003663 = Rstandard and is expressed as delta value in ‰ notation:

display math

where R denotes the 15N/14N ratio of the sample and the standard. An aliquot of 2 mg of the ground samples was then analysed for stable isotope ratio (δ15N, for community biomass also δ13) using a continuous flow stable isotope ratio mass spectrometer (Finnigan MAT DeltaPlus attached to a Carlo Erba Elementar Analysator with ConFlo II Interface). For a subset of 40 samples, duplicate measurements were performed. Among replicates, low deviations of 0.08‰ for soil and 0.03‰ for plant material suggest high representativeness of a single measurement for the respective sample.

To estimate nitrogen (N) and phosphorus (P) concentrations, biomass samples were analysed by near-infrared reflectance (NIR) spectroscopy. We recorded the reflectance spectrum of each sample between 1250 and 2350 nm at 1-nm intervals. Each sample scan consisted of 24 single measurements, which were averaged to one spectrum. Accuracy of model predictions was checked by applying an external validation process. Multiple correlation coefficients (r) and root mean square errors of cross-validation (RMSECV) of used calibration models were r = 0.99 and 0.90 and RMSECV = 0.85 and 0.45 g kg−1 for N and P, respectively. For further methodological details, see Kleinebecker, Klaus & Hölzel (2011a).

Data analysis

We analysed δ15N in the community biomass, in the soil and in the foliar biomass of three species, separately. To control for plot-specific differences in background soil δ15N values, biomass ∆δ15N values were derived plot-wise as the difference between foliar δ15N and soil δ15N. To explore the drivers of the N isotopic signals, we performed several analyses of covariance (ancova) treating the N isotopic signals as a function of a set of potentially important factors and continuous variables using the linear modelling framework in R (R Development Core Team, 2011). The independent variables of the a priori defined model were soil type (categorical variable ‘terrestrial’ versus ‘semi-terrestrial’ soil, as a stronger differentiation did not improve model fit), δ13C in single species or community biomass (as an integrative measure of water supply and potential drought stress; Adams & Grierson 2001), total grass, legume and non-legume herb cover, fertilization, grazing and mowing intensity, fertilizer type (mineral, organic, mineral and organic and no fertilizer application), study region and plant species richness. Interactions of all variables with soil type terrestrial and soil type semi-terrestrial were initially included in the models but subsequently removed if not significant. For gaining ancova results, we used the summary function to account for the co-variation among independent variables. Data were log-transformed to achieve a normal distribution of ∆δ15N and δ15N values of community biomass, Taraxacum and Dactylis. We used the Akaike Information Criterion (AIC) for model selection. Model assumptions were checked using the Shapiro–Wilk test for residuals and diagnostic plots. Furthermore, we compared means of different classes (e.g. soil types) using t-tests or the parametric Tukey honest significance test (HSD) for unequal sample size.

Results

The studied grassland plots showed large variation in vegetation and environmental characteristics and thus reflected several gradients including plant diversity and land-use intensity that all potentially affect N dynamics (Table S2). δ15N in plant biomass and soil ranged from −5.10‰ to 9.37‰ and from 0.07‰ to 6.21‰, respectively. The gradient in land-use intensity is reflected by large variation in fertilization intensity (N application rates between 0 and 225 kg ha−1 year−1), cutting frequency (0–3 cuts year−1) and grazing intensity (ranging from 0 to 1060 livestock unit days year−1). Similarly, vascular plant species richness varied from 13 to 69 species per 16 m2, and plots differed strongly in above-ground biomass production as well as in its chemical composition (e.g. N concentration of 1.2% to 3.6%, data not shown).

Plant diversity and plant functional group effects on math formula

Plant diversity was significantly related with the 15N isotopic signal. Foliar δ15N and ∆δ15N values in the community biomass and the non-legume single species samples strongly decreased with increasing plant diversity (Figs 1 and 2). ancova models for δ15N and ∆δ15N in community biomass and the two non-legume species were highly significant and revealed R2 values between 0.66 and 0.83. Except for ∆δ15N values in Dactylis, plant diversity was among the most important predictor variables affecting the 15N isotopic signal (Tables 1 and 2). Similarly, soil δ15N values significantly decreased with increasing plant diversity (Fig. 1), and, after accounting for all other predictor variables, plant diversity was the most important factor in the ancova model. No significant ancova model was derived for δ15N in Trifolium, and the respective model for ∆δ15N resulted in an R2 of 0.40 (Table 2). In contrast to the community biomass and the non-legume species, for Trifolium, a significant increase in foliar δ15N and ∆δ15N values with increasing plant diversity was observed (Fig. 2).

Table 1. Summary of analysis of covariance (ancova) of δ15N and ∆δ15N (δ15N biomass – δ15N soil) in community biomass and δ15N in soil. ‘Soil type’ consists of two soil classes: terrestrial versus semi-terrestrial soil types. ‘Fertilizer type’ consists of four classes: only mineral, only organic, mineral and organic together and no fertilizer application. Study region refers to the three regions under study (SEG = Schorfheide-Chorin). For mowing, grazing and fertilization, mean values from 2006 to 2010 were used. Arrows indicate increasing or decreasing effects of predictor variable on 15N isotopic signals. P values of predictor values < 0.05 are given in bold
Factord.f.Community biomassSoil
δ15N∆δ15Nδ15N
n Adjusted R2P model n Adjusted R2P model n Adjusted R2P model
1460.73< 0.0011460.66< 0.0011460.51< 0.001
Estimatet value P Estimatet value P Estimatet value P
Vascular plant species richness1 5.75 < 0.001 2.58 0.011 4.56 <0.001
Cover sum grasses (%)1   −0.030.975 2.71 0.008  −0.430.671
Cover sum legumes (%)1 −0.680.498 −0.940.348 0.040.968
Cover sum non-legume herbs (%)1 0.700.483 0.350.730 0.470.643
Fertilization intensity (kg*ha−1*year−1)1 2.65 0.009 2.37 0.019  1.270.208
Mowing intensity (cuts*year−1)1   −0.950.344 2.02 0.045  1.460.148
Grazing intensity (livestock units*days*year−1)1   1.070.288 0.100.920 2.02 0.046
Fertilizer type3   -0.980.327 -0.170.869 (mineral) ↓ 2.76 0.007
Above-ground N (g*sqm−1)1 3.89 < 0.001 3.49 <0.001 2.67 0.009
N:P community biomass1   −1.270.205 0.610.544 3.17 0.002
Soil type1 (terr) ↓ 2.73 0.007 (terr) ↓ 4.43 <0.001   1.280.205
Soil type:cover sum grasses1 3.00 0.003
D13C community biomass (‰)1   0.840.405 0.730.465 0.390.700
Study region2 (SEG) ↑ 2.27 0.025 (SEG) ↑ 3.36 0.001  −0.320.748
Table 2. Summary of analysis of covariance (ancova) of δ15N and ∆δ15N in single plant species. ‘Soil type’ consists of two soil classes: terrestrial versus semi-terrestrial (Gleysols, Histosols) soils. ‘Fertilizer type’ consists of four classes: only mineral, only organic, mineral and organic together and no fertilizer application. Study region refers to the three regions under study (SEG = Schorfheide-Chorin). For mowing, grazing and fertilization, mean values from 2006 to 2010 were used. Arrows indicate increasing or decreasing effects of predictor variable on 15N isotopic signals. P values of predictor values < 0.05 are given in bold
Dactylis glomerata (grass)δ15N∆δ15N
Factord.f. n Adjusted R2P model n Adjusted R2P model
590.83< 0.001590.75< 0.001
Estimatet value P Estimatet value P
Vascular plant species richness1 5.42 <0.001  −1.30.208
Cover sum grasses (%)1 −0.210.839 −1.50.144
Cover sum legumes (%)1 −0.820.417 −0.60.530
Cover sum non-legume herbs (%)1 3.10 0.003  1.50.154
Fertilization intensity (kg*ha−1*year−1)1 3.05 0.004 2.2 0.036
Mowing intensity (cuts*year−1)1 −1.680.100 −1.50.129
Grazing intensity (livestock units*days*year−1)1 2.70 0.010 3.0 0.005
Fertilizer type3 (org) ↑ 2.25 0.029 (org) ↑ 3.1 0.004
Above-ground N (g*sqm−1)1 0.010.991 −0.20.818
N:P community biomass1 −0.210.831 0.10.924
Soil type1 (terr) ↓ 3.29 0.002 (terr) ↓ 6.1 <0.001
D13C in respective species (‰)1 −0.150.882 2.00.056
Study region2 0.130.894 −1.40.174
Taraxacum sect. ruderalia (herb)δ15N∆δ15N
Factord.f. n Adjusted R2P model n Adjusted R2P model
590.82< 0.001590.73< 0.001
Estimatet value P Estimatet value P
Vascular plant species richness1 6.48 <0.001 3.52 0.001
Cover sum grasses (%)1 −0.350.728 −1.020.313
Cover sum legumes (%)1 −0.530.599 −0.150.883
Cover sum non-legume herbs (%)1 0.770.444 −0.120.906
Fertilization intensity (kg*ha−1*year−1)1 3.26 0.002 3.06 0.004
Mowing intensity (cuts*year−1)1 2.70 0.010 3.30 0.002
Grazing intensity (livestock units*days*year−1)1 0.670.509 0.490.624
Fertilizer type3 −0.900.371 0.140.892
Above-ground N (g*sqm−1)1 0.970.337 0.400.689
N:P community biomass1 −0.990.330 −1.020.315
Soil type1 (terr) ↓ 2.55 0.015 (terr) ↓ −4.35 <0.001
D13C in respective species (‰)1 1.340.189 1.740.090
Study region2 −0.440.662 −0.540.594
Trifolium repens (legume herb)δ15N∆δ15N
Factord.f. n Adjusted R2P model n Adjusted R2P model
580.907580.400.017
Estimatet value P Estimatet value P
Vascular plant species richness1 −0.120.905 2.93 0.007
Cover sum grasses1 0.620.543 −1.190.245
Cover sum legumes1 −0.730.471 0.840.411
Cover sum non-legume herbs1 −0.660.515 2.43 0.023
Fertilization intensity (kg*ha−1*year−1)1 0.940.358 −0.360.724
Mowing intensity (cuts*year−1)1 0.390.698 −0.080.937
Grazing intensity (livestock units*days*year−1)1 0.260.798 0.810.425
Fertilizer type3 −1.540.135 1.440.161
Above-ground N (g*sqm−1)1 1.030.314 −1.680.105
N:P community biomass1 0.330.745 0.330.742
Soil type1 0.830.417 (terr) ↓ 2.90 0.008
D13C in respective species (‰)1 1.270.215 0.140.887
Study region2 −0.760.455 0.140.887
Figure 1.

δ15N in community biomass (a), ∆δ15N (i.e. foliar δ15N - soil δ15N) (b) and δ15N in soil (c) as a function of plant species richness.

Figure 2.

Foliar δ15N values for (a), the herb Taraxacum sect. Ruderalia (b), the grass Dactylis glomerata (c), the legume Trifolium repens and foliar ∆δ15N values (i.e. foliar δ15N - soil δ15N) for (d) Taraxacum (e) Dactylis (f) Trifolium as function of plant species richness.

The total cover of different plant functional groups revealed relatively low and equivocal effects on biomass δ15N and ∆δ15N values (Tables 1 and 2). No significant effect on soil δ15N values was observed. Increasing herb cover increased foliar δ15N values in Dactylis and decreased foliar δ15N values Trifolium. Increasing grass cover decreased ∆δ15N values in the community biomass. Legume cover did not significantly predict any δ15N isotopic signal.

Factors beyond plant diversity

Among variables used to describe land use, fertilization intensity was clearly the most important driver of foliar δ15N and ∆δ15N values in the community biomass and the non-legume species. Irrespective of the type of fertilizer (mineral, mineral+organic, organic), higher fertilization intensity corresponded to increasing δ15N values (Tables 1 and 2). Contrarily, fertilization intensity did not significantly affect soil δ15N values, but the use of organic fertilizers led to 15N enrichment (Table 1). Grazing intensity significantly positively affected foliar δ15N and ∆δ15N values in Dactylis as well as the soil δ15N signal, whereas mowing intensity negatively affected foliar δ15N and ∆δ15N values in Taraxacum as well as ∆δ15N values in the community biomass (Tables 1 and 2). The amount of N stored in the above-ground biomass had a significant impact on δ15N and ∆δ15N values in the community biomass as well as on soil δ15N values. Increased 15N isotopic signals were consistently observed in more productive plots that stored larger amounts of N in above-ground biomass (Fig. 3). Contrastingly, δ15N in the three individual plant species was not significantly affected by the amount of N stored in the above-ground biomass (Table 2). The N:P ratio in community biomass as a measure for the relative availability of N in relation to P did not significantly affect the isotopic signal in single species and community biomass. Soil δ15N values, however, were significantly affected, becoming more depleted in 15N when the N:P ratio in the vegetation biomass was higher (Table 2).

Figure 3.

δ15N in community biomass (a), ∆δ15N (i.e. foliar δ15N - soil δ15N) (b) and soil δ15N (c) as a function of the amount of N stored in the above-ground biomass.

δ15N in soil was strongly related to δ15N in community biomass (R2 = 0.30; < 0.01) but showed no differences among single soil types (data not shown). However, the soil type did not significantly affect the soil 15N isotopic signal (Table 2). Contrary, the soil type had a significant impact on δ15N and ∆δ15N values in the community biomass and in the two non-legume species (Tables 1 and 2). Compared with terrestrial soils, δ15N isotopic values were significantly higher in biomass derived from semi-terrestrial soils (Geysols and Histosols; Fig. 4a). Biomass δ13C values (as an indicator of water supply) were negatively associated with ∆δ15N (and δ15N values) but co-varied strongly with soil type classes (Fig. 4b), thus being not significant in the ancova models (Tables 1 and 2).

Figure 4.

∆δ15N (i.e. foliar δ15N - soil δ15N) in community biomass growing on terrestrial versus semi-terrestrial soils (a) and ∆δ15N in community biomass as a function of δ13C (b). * indicates a significant difference after t-test (P < 0.05). Open dots: terrestrial soils, filled dots: semi-terrestrial soils.

The study area significantly affected the δ15N and ∆δ15N values in the community biomass, but not the 15N signals in the single species and the soil (Tables 1 and 2). Both, δ15N and ∆δ15N values were significantly higher in Schorfheide plots, even when only plots on terrestrial soils were compared (P < 0.05).

Discussion

Plant diversity and plant functional group effects on math formula

Contrary to biodiversity–ecosystem functioning experiments where the plant species richness is manipulated and other conditions are kept as constant as possible, we tested a large set of potentially important drivers of N dynamics across an established gradient of plant species richness in permanent grasslands situated in a real landscape context. δ15N signals in both plant biomass and soil revealed a clearly larger variation compared to experimental studies (Gubsch et al. 2011) and observational studies with a relatively narrow focus on specific grassland types (Kahmen et al. 2006; Beyschlag et al. 2009). In our study, increased plant diversity was found to be strongly associated with decreasing 15N isotopic signals in community biomass and the two non-legume species as well as in soil. Depletion in foliar δ15N and ∆δ15N values suggests a more complete N uptake as losses of 15N-depleted N2O and N2 or math formula via denitrification or leaching are reduced (Ometto et al. 2006; Kahmen, Wanek & Buchmann 2008). Thus, we provide strong evidence that increased plant diversity leads to a more efficient uptake and a more closed N-cycle in established semi-natural grasslands, particularly because of all the environmental noise we accounted for.

Complementarity through niche partitioning resulting in reduced interspecific competition is suggested to be the key mechanism responsible for enhanced resource-use efficiency and reduced losses under high plant diversity (Hooper et al. 2005). Thus, diverse communities are expected to take up more of the available resources as species use different resources or vary spatially and temporally in uptake patterns. Kahmen et al. (2006) state that co-occurring plant species partition the soil N pool with respect to relative resource acquisition in space, time and chemical N form. They also suggest that ecosystem N dynamics depends on plant functional group composition affiliation and is hardly driven by species-based biodiversity effects. We found a rather low and inconsistent impact of plant functional group composition on biomass 15N signals and a strong and consistent impact of plant diversity. Legumes, in particular, are considered to strongly affect N cycling due to their ability to fix atmospheric N2 (Spehn et al. 2002; Gubsch et al. 2011; Rascher et al. 2012). In our study, the cover of legumes did not significantly affect δ15N values. However, the mechanism of complementary resource utilization is not necessarily dependent on the presence of legumes as shown by van Ruijven and Berendse (2005) for artificial non-legume species combinations. Each plant species is different in its characteristics and specific traits affecting N dynamics, and thus the diversity of specific functions with which species perform within the N cycle is naturally related to the plant species richness (Scherer-Lorenzen 2005). Thus, the consistent depletion in 15N with increasing plant diversity documented here supports the hypothesis of complementary resource utilization.

Other factors driving N dynamics under natural conditions

Land use had a significant impact on δ15N and ∆δ15N in vegetation biomass and non-legume species as well as δ15N in soil. Among the land-use measures, fertilization intensity revealed the strongest and most consistent effect on the δ15N and the ∆δ15N isotopic signal in the community biomass as well as in the two non-legume species. Irrespective of the type of fertilizer, plant biomass becomes increasingly enriched in δ15N with increasing amount of fertilizer N applied. Soil processes such as denitrification and ammonia volatilization, which discriminate against 15N, were found to be stimulated by increased supply of both mineral and organic fertilizer N, leading to enhanced losses of 15N-depleted compounds (Watzka, Buchgraber & Wanek 2006). Moreover, in the studied grasslands, the impact of the fertilizer type might be masked as mineral and organic fertilizers are often applied simultaneously, and the fertilization regime type may vary among years (Blüthgen et al. 2012). Organic fertilization can additionally lead to a direct enrichment in the topsoil due to the accumulation of 15N-enriched fertilizer N (Watzka, Buchgraber & Wanek 2006). This is consistent with the positive effect of organic fertilizer application on the soil δ15N signal.

Cutting frequency appeared to have a negative impact on the 15N signal of the biomass, at least for δ15N and ∆δ15N values in Taraxacum and ∆δ15N values in community biomass. As plant biomass is usually more strongly depleted in δ15N compared to soil, frequent mowing should rather increase the 15N isotopic signal (Kriszan et al. 2009). A possible explanation for the observed pattern could be a strong correlation between cutting frequency and the amount of mineral N fertilizer (r = 0.58, P < 0.001). Thus, a decreasing effect of mineral N fertilizer on the δ15N signal particularly in intensively managed meadows might have become obvious indirectly via mowing intensity. Contrary, grazing which is often performed without the use of mineral fertilizers did not significantly affect δ15N patterns in biomass. Nevertheless, similar to the effect of organic fertilizers, grazing intensity positively affected the 15N signal of the soil reflecting the long-term effect of regular biomass removal as well as dung and urine deposition (Hoeft et al. 2012).

Except for Trifolium, plant biomass δ15N was clearly related to soil N isotopic composition as non-legume plants cover their N requirements from the soil, but in turn also affect soil δ15N by their litter which contributes to the soil N pool (Kahmen, Wanek & Buchmann 2008). Although soil types did not differ in their δ15N signal, δ15N und ∆δ15N in plant biomass were significantly affected by terrestrial versus semi-terrestrial soil conditions. Inorganic N compounds become increasingly depleted in 15N along the N cycle, thus NH4+ is usually enriched in 15N compared to math formula (Robinson 2001). Higher δ15N and ∆δ15N values in plant biomass on soils which are at least temporarily water-saturated may therefore reflect preferential NH4+ uptake (Kahmen, Wanek & Buchmann 2008). Additionally, enhanced losses of 15N-depleted mineral N via denitrification to N2O and N2 may lead to an enrichment of the remaining math formula pool. Such losses of 15N-depleted mineral N via leaching or denitrification should also lead to a gradual enrichment of 15N in the remaining soil N pool (Pardo et al. 2006; Kahmen, Wanek & Buchmann 2008). However, in our study, semi-terrestrial topsoils were not enriched in 15N, presumably due the to low contribution of the current vegetation to the long-term soil N pool. Interestingly, on semi-terrestrial soils, ∆δ15N values were not only significantly increased in the community biomass and the two non-legume species but also in Trifolium. In wet grasslands, potassium often limits primary production (Olde Venterink et al. 2003; Klaus et al. 2011) and K deficiency may also have affected biological N2 fixation by legumes resulting in higher proportions of soil-derived N in the legume biomass (Hogh-Jensen 2003).

Both the availability of N (in relation to P) and the total amount of N taken up by the vegetation affect 15N isotopic patterns in plants and soils (Hobbie & Högberg 2012). The amount of N stored in the above-ground vegetation as a proxy for total vegetation N uptake was consistently positively related to δ15N and ∆δ15N values in community biomass and the two non-legume species as well as to the soil δ15N signal. Foliar N concentrations were often found to be positively correlated with foliar δ15N values indicating high N supply due to enhanced N-turnover in the soil (e.g. Huber et al. 2007; Craine et al. 2012). The strong positive impact of vegetation N uptake on 15N isotopic signals found here indicates enhanced N fluxes (and increased losses) in more productive grassland vegetation, reflecting a more open N cycle. Although phosphorus limitation is suggested to boost fractionation against 15N (Högberg 1997; Clarkson et al. 2005), the N:P ratio in the community biomass revealed no significant effect on biomass δ15N signals. Particularly biomass P concentrations and N:P ratios were strongly related to plant diversity (r = −0.55 and = 0.25, respectively, < 0.01) as well as to the intensity of land use (Klaus et al. 2011; Blüthgen et al. 2012). Therefore, a potential effect of P deficiency on the isotopic signal could be overruled by these covarying factors. Moreover, the N:P stoichiometry is subject to considerable temporal variation (Kleinebecker, Weber & Hölzel 2011b), which might have reduced the explanatory power in this particularly year. However, the N:P ratio was a significant predictor of soil δ15N values. The mostly inert N pool in the soil integrates N dynamics over longer time periods (Johannisson & Högberg 1994), and therefore, our results reflect the longer-term effect of relative P deficiency on nitrogen dynamics independent of annual fluctuations.

Additional factors that have not been considered may have also contributed to the observed δ15N isotopic patterns. To account at least partly for those factors, we included the study region as an explanatory factor as the three regions differ in, for example, climate, land-use history and the regional species pool (Klaus et al. 2013). Additionally, mycorrhizal fungi play an important role for the N acquisition of plants. However, in permanent grasslands, most symbioses with fungi are arbuscular forms (AMF), which have low discrimination against 15N (Hobbie & Högberg 2012). Thus, mycorrhization is considered to be of minor importance for the overall N dynamics of the grasslands under study.

Conclusion

Accounting for a large set of potential drivers of ecosystem N dynamics, our results provide strong evidence that plant diversity increases nitrogen-use efficiency and reduces losses of N in real-world permanent grasslands. We also showed that nitrogen-use efficiency in permanent grasslands is decreased under more intensive land use (especially fertilization intensity), leading to enhanced losses of N to the groundwater and to the atmosphere. Thus, although single species may differ in their individual effect size, actively increasing total plant species richness in grasslands could be an option to more effectively use N resources and to reduce the negative impact of landscape scale eutrophication caused by agricultural land use.

Acknowledgements

We thank Harald Strauß and Artur Fugman for thorough isotopic analyses, Alena Jírová for help during the field campaign and Svenja Agethen and Max Moenikes for help with laboratory work. Furthermore, we thank the managers of the three exploratories: Swen Renner, Sonja Gockel, Kerstin Wiesner and Martin Gorke for their work in maintaining the plot and project infrastructure; Simone Pfeiffer and Christiane Fischer giving support through the central office, Michael Owonibi for managing the central data base and Markus Fischer, Eduard Linsenmair, Dominik Hessenmöller, Jens Nieschulze, Daniel Prati, Ingo Schöning, François Buscot, Ernst-Detlef Schulze, Wolfgang W. Weisser and the late Elisabeth Kalko for their role in setting up the Biodiversity Exploratories project. The work has been funded by the DFG Priority Program 1374 ‘Infrastructure-Biodiversity-Exploratories’ (HO 3830/2-2). Fieldwork permits were issued by the responsible state environmental offices of Baden-Württemberg, Thüringen and Brandenburg (according to § 72 BbgNatSchG).

Ancillary